Altimeter Waveform Retracking for Land and Ocean Use

1
Altimeter Waveform Retracking for Land and Ocean
Use

• Philippa Berry
• EAPRS Lab,
• De Montfort University,
• Leicester, UK

OCTAS - January 2005
2
Overview

• Fundamentals of radar satellite altimetry
• Real altimetry
• Missions
• Retracking
• Waveform shape analysis
• Ocean Results
• Applications
• Inland water
• Land mapping

3
Radar Altimetry

• Certain wavelengths of microwave energy can pass
  through the earths atmosphere with relatively
  little attenuation
• This consideration governs the frequency of
  operation of the instruments
• To date, Ku, C and S-band altimeters have been
  launched. Extensive archives of Ku band data from
  multiple instruments exist Topex gathered a
  database of C band. Envisat carries an S band
  instrument data are now being returned.

4
Atmospheric Windows
5
Why retrack waveforms?

• Most oceanographers work with pre-retracked
  waveforms in a high level product
• These data have been pre-processed using a model
  to obtain an optimal range to surface
• This model fails when waveform shapes do not
  conform to the expectation
• This can occur due to extreme sea states (high or
  low) or over non-ocean surfaces
• To work effectively with these data and to
  optimise height recovery, a knowledge of waveform
  retracking techniques is required

6
Ocean Waveforms

• Here, we overview the principles of waveform
  generation
• Then we consider basic retracking techniques
• Then proceed to look at non-ocean waveforms over
  land, inland water and ice surfaces
• Ocean data are generally well retracked by
  existing processing schemes which fit to an
  expected model
• But to obtain correct range-to-surface estimates
  over non-ocean surfaces, it is essential to
  retrack the data because the assumptions
  underlying the ocean retracking are not valid.

7
Instrument

• Pulse and Echo echo characteristics
• instrument design
• nature of surface viewed scattering mechanisms
  involved
• properties of two-way propagation path

8
Geometry of Radar Altimetry

• Range - varies due to topography, earth shape
  orbit
• Orbit - near circular
• Reference Surface - Altimeter measurements
  commonly referred to ellipsoid or geoid model.
  Note Reference Surface Range measurements MUST
  reference to a common system e.g. WGS-84

9
Range Window

• Return signals generally square-law detected, and
  resulting power profile sampled at discrete
  intervals of delay time over finite delay period
  - range window
• Typical range window around 30m geometry
  variations show orders of magnitude greater
  variation
• Thus active control is required to maintain
  return echo within range window. This is
  achieved using onboard range tracker.

10
Pulse Limited Altimetry - 1
11
Range Tracker

• Tracking point typically maintained over 50
  power point of pulse leading edge ( mean sea
  surface)
• Following loss of lock Acquisition sequence
  activated until surface signal recaptured
• Variable signal strength from surface -
  Automatic Gain Control used to maintain high
  signal to noise ratio and prevent receiver
  saturation.
• Problems - noise on individual pulses, so summed
  pulse blurring.

12
Pulse Limited Altimetry - 2

• Difference in range between antenna boresight
  edge of BLF is gt 3 ? s (max) where ? s (max) is
  maximum likely surface roughness.
• For flat, homogeneous surface, the echo power as
  function of delay time is
• P(t) PFS(t) qs(t) s r(t)
• PFS(t) is Flat Surface Impulse Response, the
  weighted variation of illuminated (flat) surface
  area as function of delay time. Shape of waveform
  leading edge is integral of height PDF of surface
  facets, and mean elevation roughness are thus
  derived from leading edge. Measurements are made
  to disk area just prior to annulus.

13
Pulse Limited Operation - 1

• For low surface roughness, diameter of Pulse
  Limited footprint is
• DPLF 2 (h c ?) 1 /2
• If roughness gt transmitted pulse length (c
  ???scattering first from crests then troughs
• Effects
• stretches leading edge (can get roughness)
• increases PLF diameter blurs edges
• DPLF 2 (h c ?) 1 /2 where ? (?2 16 ?
  s2 ln2 / c2 ) 1 /2
• For typical case
• h 800 km, ? 3ns, 1.7km lt DPLF lt 7km (at ? s
  5m)

14
Pulse Limited Operation - 2

• Range rings DRRn 2 (n h c ?) 1 /2 where n is
  range ring number (1 for PLF)
• Range window footprint
• range bins limited thus range window defines a
  footprint of diameter (ideal case)
• DRWF 2 ( h ??r) 1 /2 where ??r is range window
  width
• Typical case
• For h800 km, ??r 30m, DRWF 9.8km.
• This is true IF return is centred on range
  window.

15
Range Window

• Over the ocean, researchers need maximum
  precision
• The ocean is fairly flat
• So the range window is optimised for ocean
  returns, e.g. ERS bin widths about 45cms
• To track over topography, need more robust mode.
  So ERS Ice Mode designed with bin width about
  1.8m. Range window 4 times wider. So precision
  lost but tracking improved over non-ocean
  surfaces. Envisat has three tracking modes,
  descopes vertical precision drastically in low
  resolution mode to maintain lock over mountains

16
Other Surface Effects

• Earth curvature
• Slope Induced error ???for ??0.1 o, error
  about 1m.
• Surface Roughness
• small scale roughness (lt1cm) affects backscatter
  intensity
• large scale roughness (1 - 10m) affects leading
  edge
• Both modify polar response of surface
  backscatter.
• Surface relief - topography
• distorts waveform shape
• recorded as change in mean range for spatial
  scale gt footprint

17
Range Estimation

• Range must be corrected for atmospheric and
  instrument effects, with equation of form
• rcorr raltimeter ?rinstr ?ratmos
  ?rsurface ?rtidal
• ?rinstr Instrument corrections for instrument
  bias etc. Instrument dependent.
• ?ratmos Atmospheric path corrections given by
• ?ratmos riono rwet tropospheric rdry
  tropospheric
• ?rsurface Surface dependent. Typical ocean
  corrections include sea state bias and
  inverse barometer corrections.
• ?rtidal Both ocean land require tidal
  correction(s).

18
Instrument corrections

• Sum of internal calibration outputs, instrument
  known characteristics etc
• Usually experienced as black box fudge factors
  and added in according to mission documentation.

19
Atmospheric layers
20
Magnetosphere
can be defined as the region of space in which
the behavior of electrically charged particles is
influenced more by the planets magnetic field
than by the solar wind.
21
Sunspot Cycle multi-wavelength
22
Prominences
Actually, filaments and prominences are different
views of the same thing - cooler gas high up in
the suns atmosphere. So like sunspots, against
the bright soar surface they look dark, but
aginst the background sky they look bright.
Heres a very famous example.
23
Ionosphere
Big correction, but can be quite well modelled
except when solar event occurs. Long wavelength
in character. Dual frequency altimeters can
calculate delay directly.
24
Tropospheric correction

• Two components
• Dry tropospheric
• Wet tropospheric
• Dry tropospheric can be modelled.
• Wet tropospheric is affected on short spatial
  scales by storm fronts.
• Often have several options for these corrections,
  some direct estimates but patchy, globally may
  have to rely on models. So wet tropo events can
  get aliased into geoid.
• MUST keep same corrections to avoid anomalous
  results.

25
Range Calculation
26
Reference surfaces - EGM96 Geoid
27
Summary

• Previous current missions have flown Pulse
  Limited altimeters as.
• designed for ocean operation
• pulse limited offers advantages in terms of
  antenna design pointing requirements
• Pulse limited altimetry
• transmitted pulse propagates as part of spherical
  shell. Illuminates disk shaped surface patch
  which grows with time. Area illuminated also
  varies greatly with surface characteristics.
  Slope of leading edge of return related to
  surface roughness.Range measurement must be
  corrected for atmospheric, instrument surface
  effects.

28
Real Altimeters

• The last section showed theory.
• This section shows real data, starting with
  missions overview
• What data are required to retrack waveforms
• Then how to retrack using the OCOG technique
• Then a selection of real waveforms over ocean,
  land and ice surfaces from different missions to
  compare shapes
• Waveform sequences to see how shapes change

29
ERS-1 Mission
Of the satellites gathering waveform data, ERS-1
has the most complicated orbit strategy this
impacts on applications. In particular, the
Geodetic Mission is relevant for mapping.
Note that the 35 day phases are not spatially
coincident as the ascending nodes differ.
30
Other Missions

• ERS-2 35 day orbit pattern throughout, in phase
  with ERS-1 final 35 day orbit, with tandem
  mission phase in first year, satellites 1 day
  apart. Ice/ocean modes on both altimeters.
• Topex/Poseidon 10 day orbit pattern throughout.
  Topex dual wavelength, ocean mode only (but more
  bins than ERS)
• Jason-1 new mission. Ocean mode only.
  Near-real-time capability. Topex orbit pattern.
• Envisat new mission. Dual wavelength, 3 modes.
  ERS-2 orbit pattern. NRT capability

31
Aliasing

• Because data are actually gathered using a
  chirped pulse, then transformed into the time
  domain using Fourier transforms, aliasing occurs
  in the first and last bins as power wraps around.
• So for all instruments, a few bins have to be
  discarded prior to analysis.

32
Aliased bins
33
ERS Waveforms - 1

• The user hostile WAP (160Gbytes per year on 120
  Exabyte tapes) holds 20Hz waveform data
• Functions are available from Infoterra or ESA to
  assist in recovering data
• The structure holds 1Hz and 20Hz fields. To
  recover the waveform data, bin gain corrections
  (supplied at 1Hz) must be applied to each
  waveform in the source packet
• The tracking point is bin 32
• The ERS RA has two modes, precise mode for ocean
  operation with about 45cm wide bins

34
ERS Waveforms - 2

• Ice mode has bins 4 times wider than ocean mode.
• A hard wired mask controls the switch from ocean
  to ice mode.
• In order to protect the ocean data, the ice mask
  is generally set inside the continental land mass
• So coastal data tend to be in ocean mode, where
  much data is lost as the altimeter loses lock on
  the topography. 40 to 50 of ERS land waveforms
  captured in ocean mode contain only the leading
  edge of the waveform. Many more miss the leading
  edge and have to be discarded.

35
ERS-2 R 7 showing Land/Sea mask
36
Topex Waveform Structure

• Topex waveforms are compressed to reduce data
  storage, from 128 to 64 bins (see right)
• Tracker point bin 32
• Values have to be reconstructed from sample
  numbers
• System optimised for ocean waveforms, to maximise
  precision of range recovery

37
Reconstructing Topex data

• For ERS, all data are on the WAP tape. Data
  volumes are huge but everything is there
• For Topex, two data structures must be merged
• The GDR data containing atmospheric corrections
  and other 1Hz fields
• The SGDR containing the waveform data
• DO NOT use MGDR data over non-ocean surfaces as
  no wet tropospheric correction is present

38
Backscatter

• To work with waveform data at 20Hz (ERS), 10 Hz
  (Topex Ku) or 5 Hz (Topex C) sigma0 is sometimes
  required as well.
• ERS-1/2 sigma0 is on the WAP tapes (there is a
  correction there also)
• But Topex sigma0 is only on the GDR and at 1 Hz.
• So sigma0 must be calculated from the AGC

39
Topex Sigma0 Calculation
40
Topex Sigma0 with AGC
Using the Automatic Gain Control, this simplifies
to
41
Envisat

• For Envisat, data are available as an SGDR
  format. This contains both the GDR data and the
  waveforms, plus some limited engineering
  information.
• A tool - ENVIVIEW - enables visualisation and
  simple plots of data directly. N.B. waveform
  plotter does not work!
• SGDR data are binary, 18Hz waveforms 4
  retracking algorithms are run continually,
  designed for ocean, land ice, sea ice and a
  specific retracker for ice leading edge.
  Presence of data DOES NOT mean valid data!! No
  algorithms for land retracking are available.

42
Useful Websites

• http//icesat4.gsfc.nasa.gov/ia_home/retrack/gsfcr
  etrackdoc.960725.html
• http//wwwcpg.mssl.ucl.ac.uk/RA2_Handbook/concepts
  /ra2/ra2-mwr-PH.html

43
(No Transcript)
44
Ideal Waveform
45
Ocean Waveform and Sea State
(Picture courtesy of Rees, 1999)
46
Retracking

• Many specific retrackers exist
• Most were developed for particular applications
• The most common and robust retracker, which has
  been used for data over all surfaces, is the OCOG
  implementation

47
Retracking with OCOG

• The OCOG technique
• Recalculates an amplitude (highest power )
• Uses this to work out the width of a rectangular
  box containing the same area as that within the
  waveform curve
• Calculates the centre of gravity of the waveform

48
OCOG Algorithm
49
Thresholding

• Very often, OCOG calculation is followed by
  thresholding
• Here, the width of the waveform leading edge from
  first return to highest amplitude is calculated
• Then a specific power point is selected
• Over oceans, generally 50
• When applied over ice, generally lower (as low as
  10 )
• One of the RA-2 retrackers is an OCOG
  implementation.

50
Flags
Processing chains have flags to characterise the
echo shapes. Waveforms which fail may not be
processed. As an illustration, here are four
tests run on Topex waveforms to detect non-ocean
echoes.
51
Ice Waveform
Over non-ocean surfaces, such flags often stop
waveform processing by ocean retrackers or
incorrectly flag the data as degraded
52
Real waveforms

• The next slides show a collection of
  representative waveforms and echo sequences from
  Topex , ERS and Envisat over different surfaces
• Unlike ocean waveforms, there is wide variety in
  actual shapes
• So to characterise shape properly, thousands of
  waveforms must be examined, not just a few samples

53
Waveform plot

• Amount of returned power vs time

ERS-2
Magellan
ENVISAT
54
Waveform sequences
55
Waveform sequences

• How do the waveform shapes change along a
  satellite track