Radiative Transport Modeling of High Frequency Regional...

1
Radiative Transport Modeling of High Frequency Regional Seismograms for Event Discrimination Christopher J. Sanborn, Steven Walsh, Michele Fitzpatrick, and Vernon F. Cormier Physics Department, University of Connecticut, Storrs R EFERENCES (1) C. J. Sanborn, et al., Radiative3D http://rainbow.phys.uconn.edu/geowiki/Radiative3D (2) Cao, S., and K. J. Muirhead (1993). Finite difference modelling of Lg blockage, Geophys. J. Int. (3) Zhang, T. R., and T. Lay (1995). Why the Lg phase does not traverse oceanic crust, B.S.S.A. (4) Mendi,C. D., B.O. Ruud, and E. S. Husebye (1997). The NorthSea Lg blockage puzzle ,Geophys. J. Int. (5) Sato, H., M. C. Fehler, andT. Maeda (2012). Seismic wave propagation and scattering in the heterogeneous earth (2nd Ed.), Springer. (6) Ballard, S., J. Hipp, A. Encarnacao, C.Young, and B. Kraus (2012). A Generalized Earth Model Software Utility (GeoTess). A BSTRACT R ADIATIVE T RANSPORT The differences between earthquakes and explosions are largest in the highest recordable frequency band. In this band, scattering of elastic energy by smallscale heterogeneity (less than a wavelength) can equilibrate energy on components of motion and stabilize the behavior of the Lg wave trapped in Earth's crust. Largerscale deterministic structure (greater than a wavelength) still assumes major control over the efficiency or blockage of the Lg and the efficiency of other regional phases. We model high frequency regional seismic wave codas (24 Hz) for the combined effects of the largescale 3D (deterministic) and the small scale (statistical) structure with a radiative transport algorithm. The algorithm propagates packets of body wave energy with ray theory through a largescale deterministic 3D structure, and includes the effects of multiple scattering by smallscale statistical structure. Coda envelopes are synthesized to illustrate sensitivities to variations in the parameters describing smallscale statistical heterogeneity, intrinsic attenuation, Lg blockage due to largescale variations in crustal thickness, and the effects of tectonic release estimated from the seismograms of nuclear tests. We predict that event discriminants based on P/Lg amplitude ratios will best separate earthquake and explosion populations at frequencies 2 Hz and higher. E ARTH S TRUCTURE DETERMINISTIC STRUCTURE Examples: Changes in Moho depth Lateral variation in seismic velocity STATISTICAL STRUCTURE Example: finescale deviations of seismic velocity, due to material inhomogeneity, small cracks and fissures, etc. Random heterogeneity can be parameterized by scale length and strength parameters. From a modeling standpoint, we divide Earth structure into two categories, based on the approach used in simulation: S OFTWARE T OOL : R ADIATIVE 3D [1] FUNDED BY: AFRL Contract No. FA945315C0069, July 1, 2015 through June 30, 2018 Address correspondence to: [email protected] or [email protected] Radiative3D is a free and open source radiative transport code for synthetics generation in 3D Earth models with complex deterministic and statistical structure. Features include: Simulates earthquake and explosion radiation patterns, parameterized via momenttensor elements Radiative transport wellsuited to highfrequency synthetics Complex 3D model structure via tetrahedral grid; planned support for GeoTess model format. [6] Produces synthetic envelopes, travel time curves, or videos of energy propagation through 3D models Realistic scattering patterns in full 3D Realistic reflection/transmission handled at discontinuous interfaces, including Pwave / Swave conversion Modeling of intrinsic attenuation; separately model intrinsic vs. scattering “Q”. Homepage: http://rainbow.phys.uconn.edu/geowiki/Radiative3D C ONCLUSIONS Radiative transport is a computationally efficient method of synthesizing the very high frequency (>2.0 Hz) seismic wave field where differences between explosion and earthquake sources are largest. By incorporating both known largescale and unknown smallscale 3D structure, radiative transport can be used to predict the behavior of ratios of regional phases along specific paths, the homogenization of source radiation patterns with range, and uncertainties in traveltime picks. Future Work: Code validation: test predictions of Radiative3D against those from numerical syntheses in 3D structure. Use of Radiative3D to model chosen paths for refinement of attenuation and scattering models in regions of interest. Radiative transport is a physical modeling technique that tracks energy transport as a particle flux, using ray tracing to solve for the trajectories of millions of particles representing small quanta of elastic energy. RT is a suitable alternative to solving the full wave equation when ray theory criteria are met, and is particularly advantageous for high frequency modeling. Another advantage of radiative transport is that scattering from smallscale heterogeneity can be handled statistically, rather than requiring ultrafine model meshes which would otherwise be needed to simulate the heterogeneity deterministically. S IMULATIONS IN C RUSTAL P INCH /B ULGE M ODELS [2],[3],[4] H ETEROGENEITY AND S CATTERING M ODEL [5] Scattering Amplitudes: Scattering is treated as a stochastic process occurring on a mean-free path basis, with deflection angle and conversions determined by probability distributions: g PP (, )= l 4 4 X PP r (, ) 2 P 2l γ 0 sin 2 g PS (, )= 1 γ 0 l 4 4 X PS ψ (, ) 2 P l γ 0 q 1+ γ 2 0 - 2γ 0 cos g SP (, )= γ 0 l 4 4 X SP r (, ) 2 P l γ 0 q 1+ γ 2 0 - 2γ 0 cos g SS (, ζ )= l 4 4 X SS (, ζ ) 2 + X SS ζ (, ζ ) 2 P 2l sin 2 von Kármán Spectrum: Inhomogeneities exist at a range of scale lengths. Corner frequency determined by a. Rapid fall-off after 1/a, determined by kappa. Power spectrum affects scattering deflection angle and P/S conversion. κ = 1.0 0.5 0.3 Dependence on Parameters: Above: affect of scattering parameters on two scattering characteristics: mean free path, or average distance between scattering events, and dipole projection, which is a measure of scattering directionality (positive values indicate dominant forward scattering, negative indicates dominant backscattering.) Below left: von Kármán spectrum for various kappa values on a loglog scale. Below right: illustration of a random walk, with scattering events deflecting phonon paths from origination at source to collection at receiver. Bottom: simulated perturbation fields for various kappa values (scalelength a held fixed). Characterizing Media: Material heterogeneity treated as perturbation against locally-uniform velocity and density background Four parameters describing Scattering Media: eps: average fractional velocity perturbation size (dV/V 0 ) nu: ratio of density-perturbation to velocity-perturbation a: scale length, or auto- correlation “corner” kappa: von Kármán parameter P INCH AND B ULGE S TRUCTURES H ETEROGENEITY S PECTRUM AND C ODA P RODUCTION Pinch with Basin Pinch without Basin Crust Bulge Flat Crust Four test cases: (a) Flat crust, (b) crust pinch with sedimentary overlay and mantle upwelling, (c) crust pinch without sedimentary overlay, and (d) crust bulge protruding into mantle. Travel time curves (above) illustrate disruption of Lg and Pg energy in each pinch/bulge scenario. Energy curves (right) give insight into effects of structure. These show timeintegrated energy as a function of distance for each test case. Pinch with sedimentary basin (case b) was associated with the greatest energy reduction at long range (950 km). Bulge structure was associated with negligible energy reduction at 950 km range. Pinch structures were associated with local amplification of energy signal in the pinched region. Bulge structure was associated with local attenuation. P INCH VS . S CATTERING E FFECTS ON L G Four test cases: (a) Flat crust, (b) crust pinch with sedimentary overlay and mantel upwelling, (c) flat crust with anomalous high scattering in localized region (d) crust pinch and highscattering in pinch region. Lg effects were isolated through use of strikeslip focal mechanism and choice of Lgfavoring azimuth for seismometer array. Travel time curves illustrate disruption of Lg energy in each scenario: Each test case (b, c, and d) was associated with a reduction of energy at long range compared to baseline test case (a). Scattering structure (case c) results in substantial attenuation but does not visibly disrupt the envelope shape beyond the scattering region. Pinch structure (case b) attenuates and disrupts the envelope shape beyond the pinch region. Scattering structure (case c) results in visible backscatter in travel time curve (c). Both pinch and scatter structures were associated with local amplification of energy signal in the variation region, followed (in cases b and d) by a correction at the end of the region. Travel time curves: Color density indicates energy amplitude (squareroot of energy) as a fraction of a distancedependent reference curve determined by a powerlaw fit to the baseline (nonpinched, nonscattering) test case. Reference curve (dashed line) and timeintegrated energy (solid line) are shown as overlay plots, along with related statistics, on a 4throot scale to accommodate compressed vertical space. (Energy curves are shown in greater detail on a logarithmic scale in the Energy Comparison figure at right.) Crust variation region is outlined by dashed vertical demarcation lines. Major regional phase velocities are indicated via velocity slope lines. Energy curves: Each series represents timeintegrated energy collected at a surface seismometer as a function of distance from source event for a given test condition. The series identified as “baseline” is the flatcrust, nonscattering condition (case a). The dashed line series is a powerlaw fit to the baseline condition that serves as a colordensity reference for the travel time curves. The distance range encompassing the crust variation region is demarcated by vertical dashed lines. NonPinched + Scattering Pinched + Scattering Pinched NonPinched NonScattering High Scatter Zone (a) (b) (d) (c) Varying heterogeneity parameters in simple layered Earth models illustrates competing effects of mean free path and dipole projection measures on coda production from multiple scattering. E.g. corner scale a: (figure panel right) – Increasing a decreases mean free path, thereby increasing rate of scatter events. Increasing rate of scatter events generally increases coda production. However, increasing a also increases scattering amplitude in the forwardscattering direction, as indicated by increased dipole projection measure. (See “Dependence on Parameters,” panel above.) This means each individual scatter event is less deflectionary, reducing the cumulative effect of multiple scattering on phonon trajectory, and reducing coda production. This results in a “sweet spot” at which coda production is maximized. Sweet spot will occur in the neighborhood of a λ, where λ is the wavelength. “Short” (a = 0.1 km) “Long” (a = 12 km) “Just Right” (a = 1.0 km) Longer Corner Scale a Shorter t = 27 sec. t = 56 sec. t = 83 sec. t = 114 sec. t = 129 sec. t = 195 sec. Using Radiative3D, we simulated crust pinch, crust bulge, and enhanced heterogeneity structures in a 3D simplified crust and upper mantle model consisting of a lowvelocity sediments layer, a crust layer, a highgradient Moho transition layer, and upper mantle structure based on AK135. Crust variation zone, 100 km wide, defined between distance ranges 370 km to 470 km from source event. (Tapering for pinch/bulge structures begins at 310 km and normal thickness resumes at 530 km.) Variation zone is either: a pinch or bulge in the crust layer , a zone of anomalously high heterogeneity (scattering region), or both. Model layers follow Earthlike curvature , and include intrinsic attenuation and mild background heterogeneity. Source event is chosen as a vertical strikeslip focal mechanism in order to isolate Lg and Pg effects through azimuth selection of seismometer array. Simulations for a given model setup, source type, and frequency produce threecomponent envelopes at each seismometer location, and travel time curves for each array. 50Million phonons per run. Approx. 2–4 hours execution time, single threaded, on Intel Core i7 desktop workstation per run. Crust Pinch Model: (Profile View) Earthquake TimeSeries: Pg Lg Pg Lg Map View and Source Mechanism: Pinch/Bulge Model: The test model is fanshaped, azimuthally symmetric, and follows Earth curvature. The pinch, bulge, or highscattering region is shown as a shaded, arcshaped band. Source location is indicated by the red dot. Seismometer placement, gather area, and orientation (RTZ) is indicated by the green disc arrays. The time series below shows phonon propagation through a crust pinch Earth model and illustrates how wave fronts evolve with time. Red markers represent Pphonons and blue markers represent Sphonons. Modeling Approach: E-POSTER DOWNLOAD goo.gl/ogzFKR

Transcript of Radiative Transport Modeling of High Frequency Regional...

Page 1: Radiative Transport Modeling of High Frequency Regional ...rainbow.phys.uconn.edu/files/Sanborn_2015_AGU_Poster.pdf · Radiative Transport Modeling of High Frequency Regional Seismograms

Radiative Transport Modeling of High Frequency Regional Seismograms for Event Discrimination

Christopher J. Sanborn, Steven Walsh, Michele Fitzpatrick, and Vernon F. Cormier Physics Department, University of Connecticut, Storrs

REFERENCES

(1) C.  J.  Sanborn,  et  al.,  Radiative3Dhttp://rainbow.phys.uconn.edu/geowiki/Radiative3D  

(2) Cao,  S.,  and  K.  J.  Muirhead  (1993).  Finite  difference  modelling  of  Lg  blockage,  Geophys.  J.  Int.    

(3) Zhang,  T.  R.,  and  T.  Lay  (1995).  Why  the  Lg  phase  does  not  traverse  oceanic  crust,  B.S.S.A.  

(4) Mendi,  C.  D.,  B.  O.  Ruud,  and  E.  S.  Husebye  (1997).  The  North  Sea  Lg-­‐blockage  puzzle,  Geophys.  J.  Int.  

(5) Sato,  H.,  M.  C.  Fehler,  and  T.  Maeda  (2012).  Seismic  wave  propagation  and  scattering  in  the  heterogeneous  earth  (2nd  Ed.),  Springer.  

(6) Ballard,  S.,  J.  Hipp,  A.  Encarnacao,  C.  Young,  and  B.  Kraus  (2012).  A  Generalized  Earth  Model  Software  Utility  (GeoTess).

ABSTRACT RADIATIVE TRANSPORT

The  differences  between  earthquakes  and  explosions  are  largest  in  the  highest  recordable  frequency  band.  In  this  band,  scattering  of  elastic  energy  by  small-­‐scale  heterogeneity  (less  than   a   wavelength)   can   equilibrate   energy   on   components   of   motion   and   stabilize   the  behavior   of   the   Lg   wave   trapped   in   Earth's   crust.   Larger-­‐scale   deterministic   structure  (greater  than  a  wavelength)  still  assumes  major  control  over  the  efficiency  or  blockage  of  the   Lg   and   the   efficiency   of   other   regional   phases.  We   model   high   frequency   regional  seismic  wave  codas  (2-­‐4  Hz)  for  the  combined  effects  of  the  large-­‐scale  3-­‐D  (deterministic)  and  the  small  scale  (statistical)  structure  with  a  radiative  transport  algorithm.  The  algorithm  propagates   packets   of   body   wave   energy   with   ray   theory   through   a   large-­‐scale  deterministic  3-­‐D  structure,  and   includes   the  effects  of  multiple  scattering  by  small-­‐scale  statistical  structure.    Coda  envelopes  are  synthesized  to  illustrate  sensitivities  to  variations  in  the  parameters  describing  small-­‐scale  statistical  heterogeneity,  intrinsic  attenuation,  Lg  blockage   due   to   large-­‐scale   variations   in   crustal   thickness,   and   the   effects   of   tectonic  release   estimated   from   the   seismograms   of   nuclear   tests.   We   predict   that   event  discriminants  based  on  P/Lg  amplitude  ratios  will  best  separate  earthquake  and  explosion  populations  at  frequencies  2  Hz  and  higher. EARTH STRUCTURE

DETERMINISTIC  STRUCTURE  Examples:  • Changes  in  Moho  depth  • Lateral  variation  in  seismic  velocity

STATISTICAL  STRUCTURE  Example:  • fine-­‐scale  deviations  of  seismic  

velocity,  due  to  material  inhomogeneity,  small  cracks  and  fissures,  etc.    Random  heterogeneity  can  be  parameterized  by  scale-­‐length  and  strength  parameters.  

From  a  modeling  standpoint,  we  divide  Earth  structure  into  two  categories,  based  on  the  approach  used  in  simulation:SOFTWARE TOOL: RADIATIVE3D [1]

FUNDED  BY:    AFRL  Contract  No.  FA9453-­‐15-­‐C-­‐0069,  July  1,  2015  through  June  30,  2018Address  correspondence  to:    [email protected]  or  [email protected]

Radiative3D  is  a  free  and  open  source  radiative  transport  code  for   synthetics   generation   in   3D   Earth   models   with   complex  deterministic  and  statistical  structure.    Features  include:  

Simulates   earthquake   and   explosion   radiation   patterns,  parameterized  via  moment-­‐tensor  elements  Radiative  transport  well-­‐suited  to  high-­‐frequency  synthetics  Complex  3D  model  structure  via  tetrahedral  grid;  planned  support  for  GeoTess  model  format.[6]  Produces  synthetic  envelopes,  travel  time  curves,  or  videos  of  energy  propagation  through  3D  models  Realistic  scattering  patterns  in  full  3D  Realistic   reflection/transmission   handled   at   discontinuous  interfaces,  including  P-­‐wave  /  S-­‐wave  conversion  Modeling   of   intrinsic   attenuation;   separately   model  intrinsic  vs.  scattering  “Q”.  

Homepage:  http://rainbow.phys.uconn.edu/geowiki/Radiative3D

CONCLUSIONS

Radiative  transport  is  a  computationally  efficient  method  of  synthesizing  the  very  high  frequency  (>2.0  Hz)  seismic  wave  field  where  differences  between  explosion  and  earthquake  sources  are  largest.  By   incorporating   both   known   large-­‐scale   and   unknown  small-­‐scale  3-­‐D  structure,  radiative  transport  can  be  used  to  predict   the   behavior   of   ratios   of   regional   phases   along  specific   paths,   the   homogenization   of   source   radiation  patterns  with  range,  and  uncertainties  in  travel-­‐time  picks.  

Future  Work:  Code   validation:   test   predictions   of   Radiative3D   against  those  from  numerical  syntheses  in  3D  structure.  Use  of  Radiative3D  to  model  chosen  paths  for  refinement  of  attenuation  and  scattering  models  in  regions  of  interest.

Radiative  transport   is  a  physical  modeling  technique  that  tracks  energy  transport  as  a  particle  flux,  using  ray  tracing  to  solve  for  the  trajectories  of  millions  of  particles  representing  small  quanta  of  elastic  energy.    RT  is  a  suitable  alternative  to  solving  the  full  wave  equation  when  ray  theory  criteria   are   met,   and   is   particularly   advantageous   for   high   frequency  modeling.    Another   advantage   of   radiative   transport   is   that   scattering  from  small-­‐scale  heterogeneity  can  be  handled  statistically,  rather  than  requiring  ultra-­‐fine  model  meshes  which  would  otherwise  be  needed  to  simulate  the  heterogeneity  deterministically.

SIMULATIONS IN CRUSTAL PINCH/BULGE MODELS [2],[3],[4]

HETEROGENEITY AND SCATTERING MODEL [5]

Scattering  Amplitudes:  

• Scattering is treated as a stochastic process occurring on a mean-free path basis, with deflection angle and conversions determined by probability distributions:

gPP (⌅, ⇥) =l4

4⇤

��XPPr (⌅, ⇥)

��2 P✓2l

�0sin

2

gPS(⌅, ⇥) =

1

�0

l4

4⇤

��XPS� (⌅, ⇥)

��2 P✓

l

�0

q1 + �2

0 � 2�0 cos⌅

gSP(⌅, ⇥) = �0

l4

4⇤

��XSPr (⌅, ⇥)

��2 P✓

l

�0

q1 + �2

0 � 2�0 cos⌅

gSS(⇤, �) =l4

4⇥

⇣��XSS⇥ (⇤, �)

��2 +��XSS

� (⇤, �)��2⌘P

✓2l sin

2

von  Kármán  Spectrum:  

• Inhomogeneities exist at a range of scale lengths.

• Corner frequency determined by a.

• Rapid fall-off after 1/a, determined by kappa.

• Power spectrum affects scattering deflection angle and P/S conversion.

κ = 1.0 0.5 0.3

Dependence  on  Parameters:

Above:   affect   of   scattering   parameters   on   two   scattering   characteristics:   mean   free   path,   or   average  distance  between   scattering   events,   and  dipole   projection,  which   is   a  measure   of   scattering  directionality  (positive  values  indicate  dominant  forward  scattering,  negative  indicates  dominant  back-­‐scattering.)  Below  left:  von  Kármán  spectrum  for  various  kappa  values  on  a  log-­‐log  scale.  Below  right:  illustration  of  a  random  walk,   with   scattering   events   deflecting   phonon   paths   from   origination   at   source   to   collection   at   receiver.    Bottom:  simulated  perturbation  fields  for  various  kappa  values  (scale-­‐length  a  held  fixed).  

Characterizing  Media:  

• Material heterogeneity treated as perturbation against locally-uniform velocity and density background

• Four parameters describing Scattering Media:

• eps: average fractional velocity perturbation size (dV/V0)

• nu: ratio of density-perturbation to velocity-perturbation

• a: scale length, or auto-correlation “corner”

• kappa: von Kármán parameter

PINCH AND BULGE STRUCTURES

HETEROGENEITY SPECTRUM AND CODA PRODUCTION

Pinch  with  Basin Pinch  without  Basin Crust  BulgeFlat  Crust

Four  test  cases:  (a)  Flat  crust,  (b)  crust  pinch  with  sedimentary  overlay  and  mantle  upwelling,  (c)  crust  pinch  without  sedimentary  overlay,  and  (d)  crust  bulge  protruding  into  mantle.  

Travel  time  curves  (above)  illustrate  disruption  of  Lg  and  Pg  energy  in  each  pinch/bulge  scenario.  

Energy  curves  (right)  give  insight  into  effects  of  structure.    These  show  time-­‐integrated  energy  as  a  function  of  distance  for  each  test  case.  

Pinch   with   sedimentary   basin   (case   b)   was   associated   with   the  greatest  energy  reduction  at  long  range  (950  km).  

Bulge   structure   was   associated   with   negligible   energy   reduction   at  950  km  range.  

Pinch  structures  were  associated  with   local  amplification  of  energy  signal  in  the  pinched  region.  Bulge  structure  was  associated  with  local  attenuation.

PINCH VS. SCATTERING EFFECTS ON LG

Four  test  cases:  (a)  Flat  crust,  (b)  crust  pinch  with  sedimentary  overlay  and  mantel  upwelling,  (c)  flat  crust  with  anomalous  high  scattering  in  localized  region  (d)  crust  pinch  and  high-­‐scattering  in  pinch  region.  

Lg   effects  were   isolated   through   use   of   strike-­‐slip   focal  mechanism  and  choice  of  Lg-­‐favoring  azimuth  for  seismometer  array.  

Travel  time  curves  illustrate  disruption  of  Lg  energy  in  each  scenario:

Each  test  case  (b,  c,  and  d)  was  associated  with  a  reduction  of  energy  at  long  range  compared  to  baseline  test  case  (a).  

Scattering  structure  (case  c)  results  in  substantial  attenuation  but   does   not   visibly   disrupt   the   envelope   shape   beyond   the  scattering  region.  

Pinch  structure  (case  b)  attenuates  and  disrupts  the  envelope  shape  beyond  the  pinch  region.  

Scattering  structure   (case  c)  results   in  visible  back-­‐scatter   in  travel  time  curve  (c).  

Both  pinch  and  scatter  structures  were  associated  with  local  amplification  of  energy  signal  in  the  variation  region,  followed  (in  cases  b  and  d)  by  a  correction  at  the  end  of  the  region.

Travel  time  curves:  Color  density  indicates  energy  amplitude  (square-­‐root  of  energy)  as  a  fraction  of  a  distance-­‐dependent  reference  curve  determined  by  a  power-­‐law  fit  to  the  baseline  (non-­‐pinched,  non-­‐scattering)  test  case.    Reference  curve  (dashed  line)  and  time-­‐integrated    energy  (solid  line)  are  shown   as   overlay   plots,   along   with   related   statistics,   on   a   4th-­‐root   scale   to   accommodate  compressed  vertical  space.    (Energy  curves  are  shown  in  greater  detail  on  a  logarithmic  scale  in  the  Energy  Comparison  figure  at  right.)    Crust  variation  region  is  outlined  by  dashed  vertical  demarcation  lines.    Major  regional  phase  velocities  are  indicated  via  velocity  slope  lines.  

Energy   curves:    Each   series   represents   time-­‐integrated   energy   collected   at   a   surface  seismometer  as  a  function  of  distance  from  source  event  for  a  given  test  condition.    The  series   identified   as   “baseline”   is   the   flat-­‐crust,   non-­‐scattering   condition   (case   a).    The  dashed  line  series  is  a  power-­‐law  fit  to  the  baseline  condition  that  serves  as  a  color-­‐density  reference  for  the  travel  time  curves.    The  distance  range  encompassing  the  crust  variation  region  is  demarcated  by  vertical  dashed  lines.

Non-­‐Pinched  +  Scattering Pinched  +  Scattering

PinchedNon-­‐Pinched

Non

-­‐Sca

ttering

High  Sc

atter  Z

one

(a) (b)

(d)(c)

Varying  heterogeneity  parameters  in  simple  layered  Earth  models  illustrates  competing  effects  of  mean  free  path  and  dipole   projection   measures   on   coda   production   from  multiple  scattering.  

E.g.   corner   scale   a:   (figure   panel   right)   –   Increasing   a  decreases   mean   free   path,   thereby   increasing   rate   of  scatter  events.    

Increasing   rate   of   scatter   events   generally   increases   coda  production.  However,  increasing  a  also  increases  scattering  amplitude  in  the  forward-­‐scattering  direction,  as  indicated  by  increased  dipole  projection  measure.    (See  “Dependence  on  Parameters,”  panel  above.)  

This   means   each   individual   scatter   event   is   less  deflectionary,   reducing   the   cumulative   effect   of   multiple  scattering   on   phonon   trajectory,   and   reducing   coda  production.  

This   results   in   a   “sweet   spot”   at  which   coda   production   is  maximized.    Sweet   spot  will  occur   in   the  neighborhood  of  a  ≈  λ,  where  λ  is  the  wavelength.

“Short”(a = 0.1 km)

“Long”(a = 12 km)

“Just Right”(a = 1.0 km)

Long

er                                        Co

rner    S

cale    a                                        Sh

orter

t  =  27  sec. t  =  56  sec. t  =  83  sec. t  =  114  sec. t  =  129  sec. t  =  195  sec.

Using  Radiative3D,  we  simulated  crust  pinch,  crust  bulge,  and  enhanced  heterogeneity  structures   in  a  3D  simplified  crust  and  upper  mantle  model  consisting  of  a   low-­‐velocity   sediments   layer,  a   crust   layer,  a  high-­‐gradient  Moho  transition  layer,  and  upper  mantle  structure  based  on  AK-­‐135.  

Crust   variation   zone,   100   km   wide,   defined   between   distance   ranges  370  km  to  470  km  from  source  event.    (Tapering  for  pinch/bulge  structures  begins  at  310  km  and  normal  thickness  resumes  at  530  km.)  

Variation   zone   is   either:   a   pinch   or   bulge   in   the   crust   layer,   a   zone   of  anomalously  high  heterogeneity  (scattering  region),  or  both.  

Model  layers  follow  Earth-­‐like  curvature,  and  include  intrinsic  attenuation  and  mild  background  heterogeneity.  

Source  event  is  chosen  as  a  vertical  strike-­‐slip  focal  mechanism  in  order  to  isolate  Lg  and  Pg  effects  through  azimuth  selection  of  seismometer  array.  

Simulations   for  a  given  model   setup,   source   type,  and   frequency  produce  three-­‐component   envelopes   at   each   seismometer   location,   and   travel  time  curves  for  each  array.  

50-­‐Million   phonons   per   run.     Approx.   2–4   hours   execution   time,   single-­‐threaded,  on  Intel  Core  i7  desktop  workstation  per  run.

Crust  Pinch  Model:  (Profile  View)

Earthquake  Time-­‐Series:

Pg

Lg

Pg

Lg

Map  View  and  Source  Mechanism:

Pinch/Bulge  Model:  The   test   model   is   fan-­‐shaped,  azimuthally   symmetric,   and   follows  Earth   curvature.   The   pinch,   bulge,   or  high-­‐scattering   region   is   shown   as   a  shaded,  arc-­‐shaped  band.    Source   location   is  indicated   by   the   red   dot.   Seismometer  placement,   gather   area,   and   orientation   (RTZ)   is  indicated  by  the  green  disc  arrays.

The  time  series  below  shows  phonon  propagation  through  a  crust  pinch  Earth  model  and  illustrates  how  wave  fronts  evolve  with  time.    Red  markers  represent  P-­‐phonons  and  blue  markers  represent  S-­‐phonons.

Modeling  Approach:

E-POSTER DOWNLOAD goo.gl/ogzFKR