onboard(processor( - nasa swot - home · 2017-03-07 ·...

23
Onboard Processor Gerald Dibarboure, Tom Farrar, Sarah Gille, Rosemary Morrow, and others

Upload: others

Post on 22-Apr-2020

2 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Onboard(Processor( - NASA SWOT - Home · 2017-03-07 · Onboard(Processor(Gerald(Dibarboure,(Tom(Farrar,(Sarah(Gille,(Rosemary(Morrow,(and(others

Onboard  Processor  

Gerald  Dibarboure,  Tom  Farrar,  Sarah  Gille,  Rosemary  Morrow,  and  others  

Page 2: Onboard(Processor( - NASA SWOT - Home · 2017-03-07 · Onboard(Processor(Gerald(Dibarboure,(Tom(Farrar,(Sarah(Gille,(Rosemary(Morrow,(and(others

Outline  

•  Introduc>on:      •  The  OBP  and  the  ATBD:    The  Basics.  •  Tes>ng  the  Onboard  Processor.    What  extreme  condi>ons  does  it  need  to  handle?  

•  Resolu>on:    250  m  vs  1  km.    What  does  the  SDT  advise?  

Page 3: Onboard(Processor( - NASA SWOT - Home · 2017-03-07 · Onboard(Processor(Gerald(Dibarboure,(Tom(Farrar,(Sarah(Gille,(Rosemary(Morrow,(and(others

Introduc>on:    OBP  Decisions  to  be  Made  in  2015  

•  Azimuth-­‐range  vs  geographic  grid  averaging    •  OBP  valida>on  (sea  state  bias)    •  Resolu>on  (1  km,  0.5  km  at  250  m  pos>ngs?)  trade-­‐offs  including  synergis>c  science    

Page 4: Onboard(Processor( - NASA SWOT - Home · 2017-03-07 · Onboard(Processor(Gerald(Dibarboure,(Tom(Farrar,(Sarah(Gille,(Rosemary(Morrow,(and(others

Onboard  Processor  (OBP)  Issues  OBP  is  a  hardware  implementa>on  q   it  receives  the  Karin  radar  echos  q   it  produces  at  1x1  km  resolu>on  &  1  km  pos>ng  :    

–  the  complex  interferogramme  product    –  power  SAR  images  for  each  of  the  2  input  channels  

 An  alternate  fixed  resolu>on  500mx500m  product  with  250m  pos>ng  is  also  being  considered,  and  addi>onal  sta>s>cs  (power  image  at  250  m  resolu>on,  sum  of  squares  at  250  m  resolu>on,  wave  mi>ga>on  algorithms,  …)      Workplan  to  be  presented/discussed  here,  at  Jan  2015  SDT  mee?ng    Results  to  be  presented/discussed  at  June  2015  SDT  mee>ng    Decision  to  be  made  by  Oct  2015  for  hardware  implementa>on    

1  km   250  m  

Page 5: Onboard(Processor( - NASA SWOT - Home · 2017-03-07 · Onboard(Processor(Gerald(Dibarboure,(Tom(Farrar,(Sarah(Gille,(Rosemary(Morrow,(and(others

What  everyone  needs  to  know  about  the  OBP  and  the  ATBD  (Algorithm  

Theore>cal  Basis  Document)  

•  OBP  will  transmit  interferograms-­‐-­‐-­‐not  ssh,  and  not  raw  data  

•  Nadir-­‐al>metry-­‐style  reprocessing  is  not  an  op>on.  •  Sea  ice  or  land  could  corrupt  full  1-­‐km  or  500-­‐m  pixel  •  JPL  technical  reports  on  ATBD  cover  a  broad  range  of  details:    sur]oard,  use  of  the  mean  sea  surface  in  OBP,  mul>-­‐look  averaging,  wave-­‐mi>ga>on  sta>s>cs.    

Page 6: Onboard(Processor( - NASA SWOT - Home · 2017-03-07 · Onboard(Processor(Gerald(Dibarboure,(Tom(Farrar,(Sarah(Gille,(Rosemary(Morrow,(and(others

Gridding  of  data  

Page 7: Onboard(Processor( - NASA SWOT - Home · 2017-03-07 · Onboard(Processor(Gerald(Dibarboure,(Tom(Farrar,(Sarah(Gille,(Rosemary(Morrow,(and(others

Gridding  of  data  

•  Convenient  for  aligning  ascending  and  descending.  

• Matches  satellite  sampling  •  If  nadir  track  shi_s,  can  debate  whether  to  use  fixed  or  >me-­‐varying  reference  track.  

Page 8: Onboard(Processor( - NASA SWOT - Home · 2017-03-07 · Onboard(Processor(Gerald(Dibarboure,(Tom(Farrar,(Sarah(Gille,(Rosemary(Morrow,(and(others

Gridding  of  data  

•  BUT:    There  is  no  na>ve  grid.    OBP  output  for  each  of  9  beams  is  not  aligned,  so  all  output  interpolated  onto  final  grid.    

Page 9: Onboard(Processor( - NASA SWOT - Home · 2017-03-07 · Onboard(Processor(Gerald(Dibarboure,(Tom(Farrar,(Sarah(Gille,(Rosemary(Morrow,(and(others

Some  OBP  Issues  …  1  Impact  of  wave  field  •  workplan  underway  with  ADT  &  Project,  &  SDT  PIs  (eg  F.  Ardhuin;  B.  Chapron)  •  Tes>ng  impact  of  different  sea  state  (wave,  swell)  regimes  on  simulated  error  •  See  following  presenta>on  …  

Example  :  2  m  wave  spectra  near  Hawaii  

Page 10: Onboard(Processor( - NASA SWOT - Home · 2017-03-07 · Onboard(Processor(Gerald(Dibarboure,(Tom(Farrar,(Sarah(Gille,(Rosemary(Morrow,(and(others

Significant  wave  height  

(Project  specifica>ons:    dra_ed  assuming  2  m  average  swh  and  5.5  m  extreme.)  

Page 11: Onboard(Processor( - NASA SWOT - Home · 2017-03-07 · Onboard(Processor(Gerald(Dibarboure,(Tom(Farrar,(Sarah(Gille,(Rosemary(Morrow,(and(others

Discussion:    Tes>ng  the  Onboard  Processor  

•  High  significant  wave  scenarios  under  considera>on-­‐-­‐-­‐what  might  be  missed?  

•  Sea  ice,  moving  objects,  slicks  not  fully  considered  yet.    What  is  needed?  

•  Are  addi>onal  cases  needed?  

Page 12: Onboard(Processor( - NASA SWOT - Home · 2017-03-07 · Onboard(Processor(Gerald(Dibarboure,(Tom(Farrar,(Sarah(Gille,(Rosemary(Morrow,(and(others

Resolu>on:  500  m  (with  250  m  pos>ngs)  vs  1  km  vs  ???  

•  Oceanographic  signals  in  the  500  m  to  5  km  range?    

•  Will  250-­‐m  pos>ngs  actually  help  us  work  around  ice  and  coastlines,  or  ships?      

Page 13: Onboard(Processor( - NASA SWOT - Home · 2017-03-07 · Onboard(Processor(Gerald(Dibarboure,(Tom(Farrar,(Sarah(Gille,(Rosemary(Morrow,(and(others

What  will  SWOT  SSH  look  like?  

Rocha  et  al,  in  prep  for  JPO,  2015  

Page 14: Onboard(Processor( - NASA SWOT - Home · 2017-03-07 · Onboard(Processor(Gerald(Dibarboure,(Tom(Farrar,(Sarah(Gille,(Rosemary(Morrow,(and(others

Some  OBP  Issues  …  2  Impact  of  noise  &  anomalous  points  in  field  (following  from  HR  mask  discussion)    These  include  ….    Permanent  structures  :  -­‐  coasts,  fords,  con>nental  

ice,  oil  plagorms,  wind  farms,  coral  reefs,  …  

Moving  structures  :  -­‐  sea-­‐ice,  icebergs,  ships,  >dal  

sandbars,  …  

Analysis  SST  field  –  E.  Autret,  PhD  Thesis    (2  lines  :  FFT  &  Discrete  Cosine  Transform)  

+Random  noise  

+Anomalous  pts  

Simulated  SST  

Page 15: Onboard(Processor( - NASA SWOT - Home · 2017-03-07 · Onboard(Processor(Gerald(Dibarboure,(Tom(Farrar,(Sarah(Gille,(Rosemary(Morrow,(and(others

•  Jason-­‐class  al>meters  give  average  parameters  over  a  large  disc  (e.g.  7-­‐10km)  •  Ar>fact  is  caused  by  approxima>ons  in  processing  algorithms  (Brown  models)  •  Triggered  by  geophysical  causes  (e.g.  atmosphere  and  sea  state  transi>ons)  •  The  hump  ar>fact  appears  when  the  al>meter  footprint  repeatedly    samples  

inhomogenei>es  in  sigma0  or  SWH  

 

Inhomogeni?es  also  impact  on  nadir  al?meter  «  noise  »  

Inner  footprint  leading  edge  

Outer  footprint  trailing  edge  

Page 16: Onboard(Processor( - NASA SWOT - Home · 2017-03-07 · Onboard(Processor(Gerald(Dibarboure,(Tom(Farrar,(Sarah(Gille,(Rosemary(Morrow,(and(others

Using  KaRIN  imagery  to  understand  1D  al>metry  

•  Karin  will  provide  a  2D  image  of  what  tradi>onal  al>meters  have  been  integra>ng  in  their  footprint  and  waveforms    è  Higher  KaRIN  resolu>on  is  desirable  for  such  studies  

•  KaRIN  imagery  may  help  inves>gate  what  happens  in  «  hump  genera>on  events  »  and  improve  decades  of  small  scale  al>metry  records  

•  Implica?on:  We  need  to  collect  reference  datasets  from  KaRIN    or  AirSWOT  with  high  resolu>on  and  nadir  overlaps  

LRM

SARM

Karin image (1 km)

LRM

SARM

Karin image (250 m)

Page 17: Onboard(Processor( - NASA SWOT - Home · 2017-03-07 · Onboard(Processor(Gerald(Dibarboure,(Tom(Farrar,(Sarah(Gille,(Rosemary(Morrow,(and(others

Hump  ar>fact  vs  Coverage  •  Hump  mi>gated  by  2  to  4  if  we  sacrifice  5-­‐15%  of  SSH  coverage  •  This  method  edits  out  major  events  (e.g.  rain  cells,  σ0  blooms)  •  Limit:  the  ar>fact  never  disappears  even  if  we  discard  50%  of  data  =>  SWOT  –  cannot  re-­‐edit/re-­‐process  the  1  km  OBP  data  !  

1  Hz  Edi>ng  Labroue  et  al:    10%  edited  out  

CWT:15%  edited  out  

CWT:  50%  edited  out  

SARAL  /  Al>Ka  –  SSHA  Power  Spectrum  

Page 18: Onboard(Processor( - NASA SWOT - Home · 2017-03-07 · Onboard(Processor(Gerald(Dibarboure,(Tom(Farrar,(Sarah(Gille,(Rosemary(Morrow,(and(others

Discussion:    OBP  Resolu>on  

•  Having  the  250  m  image  will  help  us  iden>fy  noise  or  anomalous  points  in  the  1  km  SSH  image;  we  can’t  reprocess  the  OBP  SSH  field  later,  to  correct  anomalies  

•  Can  we  develop  sta>s>cal  techniques  within  the  OBP  to  remove  /  reduce  anomalous  points  in  1  km  or  500  m  SSH  fields  ?  

•  Can  SDT  members  provide  test  cases  of  images  /  heights  with  anomalous  points  to  the  Project,  to  test  with  their  simulator?  

1  km   250  m  

 4  x  1  km  

 #  Nadir  al>meter  waveform  perturba>ons  due  to  ships,  over  4  mon  on  1x2°  grid  

Tournadre,  GRL,  2014.  

Page 19: Onboard(Processor( - NASA SWOT - Home · 2017-03-07 · Onboard(Processor(Gerald(Dibarboure,(Tom(Farrar,(Sarah(Gille,(Rosemary(Morrow,(and(others

Extra  Slides  

Page 20: Onboard(Processor( - NASA SWOT - Home · 2017-03-07 · Onboard(Processor(Gerald(Dibarboure,(Tom(Farrar,(Sarah(Gille,(Rosemary(Morrow,(and(others

Gridding  and  image  distor>on  

Page 21: Onboard(Processor( - NASA SWOT - Home · 2017-03-07 · Onboard(Processor(Gerald(Dibarboure,(Tom(Farrar,(Sarah(Gille,(Rosemary(Morrow,(and(others

Noise  and  spurious  pixels  

Unfiltered   Filtered:  spurious  signals  can  be  harder  to  isolate  

Page 22: Onboard(Processor( - NASA SWOT - Home · 2017-03-07 · Onboard(Processor(Gerald(Dibarboure,(Tom(Farrar,(Sarah(Gille,(Rosemary(Morrow,(and(others

Blurring  of  direc>onal  signals  

Page 23: Onboard(Processor( - NASA SWOT - Home · 2017-03-07 · Onboard(Processor(Gerald(Dibarboure,(Tom(Farrar,(Sarah(Gille,(Rosemary(Morrow,(and(others

Moiré  effects:    original  data  grid  not  aligned  with  mapping  grid