Unifying multimission altimeter sigma0 a new approach

1
Unifying multi-mission altimeter sigma0 a new
approach

• Susan M.S. Bramer, Philippa A.M. Berry, Jennifer
  A. Freeman and E. Lucy Mathers

2
Contents

• Introduction
• Expert system
• Finding natural land calibration targets
• Modelling calibration zones with ERS-1 GM
• Cross-calibration of ERS-1/2 and Envisat sigma0
• Automation of model creation
• Cross-calibration with TRMM
• Jason data - first results
• Future - building a TOPEX model

3
Introduction

• To cross calibrate altimeter sigma0 over ocean
  co-temporal, co-spatial data required.
• This is not always possible
• Another approach is to use natural land
  calibration targets to cross calibrate.
• In order to do this detailed sigma0 models of the
  land calibration targets must be created.

4
Introduction

• Detailed sigma0 modelling is required over land
  because sigma0 response varies widely over land,
  even over 2km dead band (Berry et al 2005).
• Models manually created for 4 desert areas
  (Johnson 2002).
• Echo shapes complex and highly variable over
  land. Use expert system approach to analyse and
  classify individual echoes.

5
Expert System

• Altimeter return waveforms over land do not
  conform to Brown Model
• Expert System interprets complex waveforms and
  chooses suitable retracker.

6
Model Creation

• Global search undertaken for regions stable in
  sigma0 response.
• Requirement sigma0 values near crossovers vary
  by less than 1dB.
• As Geodetic Mission tracks time separated, only
  areas with little environmental contamination
  appear coherent when plotted.
• Detailed Crossover Analysis employed.
• Model overpopulated along 35 Day Mission tracks.

7
Simpson Desert Model (Johnson 2002)

• Environmental Contamination removed manually.
• Tuned for ERS-1/2.
• We require detailed modelling of ground tracks of
  other satellites, other areas.
• Technique unsuitable for wetter areas.

• Mean values taken where multiples.
• Delauney Triangulation and Bilinear Interpolation
  create surfaces.

8
Model Testing

• Test tracks excluded from model and compared with
  predictions.

Test Track over Simpson Desert (Johnson
2002) Top red actual values, blue modelled
values. Bottom differences.
9
Model Usage

• The models were then used for cross calibration
  of ERS-1/2 ice/ocean mode sigma0 (Johnson 2002).
• ERS-1 to ERS-2 ice mode cross calibration offset
  7.55 0.02 dB
• ERS-1 Ice to Ocean mode 9.06 0.01 dB
• ERS-1 ice to ERS-2 ocean 7.24 0.14 dB
• These models have also been used to study the
  validity of the Envisat SGDR values over land,
  and retracked Sigma0.

10
Model Usage
Retracked Envisat Ku and S Band with Sigma0 Model
11
Model Usage

• BUT manual model building prohibitively time
  consuming and subjective.

12
Automation of Model Production

• Software tools developed to enable automated
  removal of environmental contamination.
• Only contaminated sections of track removed.

• Crossover Analysis used to reconcile 35 day
  tracks to Geodetic Mission.
• Minimum values always preferred as representing
  dryer conditions.

Cleaned (green) and original, environmentally
contaminated (red) 35 day repeat tracks
13
Original and New Simpson Desert Models, ERS-1/2

• New model and original have correlation
  coefficient 0.935
• Original model takes mean values, new model takes
  minimum.
• It is not expected that new models will be quite
  as smooth as old, BUT

Huge savings in time, and some data extra data
kept
14
Cross Calibration with TRMM PR

• Can we connect this modelled sigma0 to the TRMM
  dataset sigma0 to prove this cross-calibration
  even works between different sensors?
• PR 2A21 surface Sigma0 product processed and
  compared with model.

15
Cross Calibration with TRMM PR

• Correlation between altimeter and PR decreases
  rapidly as PR swings away from nadir.
• PR more sensitive to vegetation as moves away
  from nadir.
• Higher correlation is obtained over Eastern
  Sahara region (less vegetation)

16
Jason/TOPEX cross calibration

• So technique works for cross-calibration.
• Wish to use for TOPEX.
• But need to overpopulate model along ground
  tracks, and questions over TOPEX sigma0 mean
  cannot be used.
• So look to Jason-1.

17
Jason-1 land sigma0 A first look

• Some Jason-1 echoes now processed through expert
  system and compared with model.
• Initial results show reasonable agreement along
  tracks crossing central Simpson Desert, noting
  that model is reliant on GM only along Jason
  tracks.

• Mean Difference from Model (7 Cycles) -3.084
• Mean difference is greater if only central
  calibration zone used.
• Mean Correlation Coefficient 0.772
• Standard deviation of track correlation
  coefficients 0.053
• Assimilating Jason will improve the model
  detailing.
• Detailed statistics will be calculated prior to
  assimilation into model.

18
Jason-1 land sigma0 A first look

• Jason repeat track 88 and Sigma0 model (red).

19
Next steps

• Now populating model with 1 year of Jason-1 data.
• Then will fly entire TOPEX mission over
  calibration models to assess sigma0 stability,
  will solve for corrections and apply over ocean,
  then compare results with TOPEX GDR values.

20
Conclusion

• Cross calibration of sigma0 response of different
  altimeters is very successful over natural land
  targets.
• In order to achieve this detailed sigma0 models
  must be created.
• For best results models must be over-populated
  along ground tracks, hence Jason values may be
  used to validate TOPEX.
• Over land correlations between such differing
  instruments as an altimeter and TRMM PR may also
  be studied.