evaluation of the ground retracking algorithms on jason-1 data

Evaluation of the ground retracking algorithms on

Jason-1 data P.Thibaut, S. LabroueCollecte Localisation Satellite : Toulouse,

France

Hobart OSTST MeetingPierre THIBAUT – March 2007

Objective

The objective of this presentation is to characterize the various retracking

algorithms applied on Jason and Topex waveforms and to determine if they can be

considered as responsible for differences observed between Jason and Topex

SSB.

We decided to apply the various retracking on Jason-1 WFs. If no differences

are observed when retracking the same waveforms, the differences between

Topex and Jason SSB may directly come from the WFs (leakages on Topex WFs) All the results presented have been obtained on one cycle of Jason-1 data (cycle 20)

• GDR ‘B’ nominal Jason-1 products• Retracked files provided by JLP (same information than in Topex RGDRs including LSE and MAP estimations)


Characteristics of the 3 retracking algorithms

MLE4(nominal proc.for GDRs) :

LSE

MAP

• Epoch, SWH, Pu, slope of the trailing edge : 20 estimations per second• Retracking a single gaussian + LUT which provide the correction between one gaussian and the full real PTR (as a function of SWH)• Skewness s set to 0.1

• Waveforms averaging (2x2 in Ku band, 4x4 in C band) before retracking• Decomposition of the PTR into a sum of Gaussians (for 20 side-lobes)• Epoch : 10 estimations per second in Ku (5 in C)• SWH, Pu, slope of the trailing edge : 1 estimation per second• s : 1 estimation per second or fixed to 0

• Idem LSE

Allows comparisonto Topex LSE

Allows comparisonto Jason MLE4

(except that for Jason a 0.1 value is used)


Range1Hz = TrackerGDR – EpochJPL – LUTGDR – RangeGDR_1Hz

Information provided by the various products

JPL products provide:the retracked epoch (10 Hz and 1 Hz) (JPL = LSE or MAP) : EpochJPL

GDR products provide:the range (20 Hz and 1Hz) : RangeGDR

the tracker range (20Hz and 1 Hz) : TrackerGDR

the Look Up Table correction for epoch (1Hz) : LUTGDR

Comparison between JPL and GDR ranges is given by:

Range1Hz = RangeJPL_1Hz – RangeGDR_1Hz

But we have only the EpochJPL


Delta Range(LSE - GDR) versus (SWH and SIG0)

40 cm 50 cm

No remaining dependancies with SWH or SIG0 in the bulk of the data

Skewness solved Range_LSE-Range_GDR versus (SWH,SIG0)

Skew solved


Delta Range(LSE - GDR) versus (SWH and ATT2)

40 cm 50 cm

No remaining dependancies with SWH or ATT2 in the bulk of the data

Range_LSE-Range_GDR versus (SWH,ATT)

Skew solved

Skewness solved


Delta Range(LSE - GDR) versus (SWH and SIG0 and ATT2)

Dependances appear when the skewness is fixed but it was fixed to 0 (in GDR 0.1)

Range_LSE-Range_GDR versus (SWH,ATT)

Skew fixed

Range_LSE-Range_GDR versus (SWH,SIG0)

Skew fixed

40 cm 50 cm 40 cm 50 cm


SWH

LSE-GDR

MAP-GDR

SWH

ATT2

ATT2

Range

Range

Very good agreement between LSE and MLE4

MAP introduces SWH and ATT dependances

(Constraints on MAP skewness is too strong : s=0)

Skewness solved

45 cm

50 cm

45 cm

50 cm

1%SWH


Delta Range(MAP - GDR) versus (SWH and SIG0 and ATT2)

MAP introduces dependances but the skewness remains close to 0 (in GDR 0.1) Same kind of plot when skewness fixed at 0

Range_MAP-Range_GDR versus (SWH,ATT) Skew solved

Range_MAP-Range_GDR versus (SWH,SIG0) Skew solved

40 cm 50 cm 40 cm 50 cm


Difference of range (LSE – MAP)

-5cm +5cm


Skewness estimation

LSE MAP

Skewness is not estimated for SWH smaller than 1 m Skewness remains close to 0 for MAP

Mean=0.06 Mean=0.001


Skewness with respect to SWH

LSE MAP

Skewness is not estimated for SWH less than 1 m Skewness remains close to 0 for MAP


ATT_LSE ATT_MAPATT_GDR

ATT(LSE-GDR) ATT(MAP-GDR)

Statistics on ATT – Skewness Solved

Mean=-0.002 Mean=-0.003

Mean=0.003 Mean=0.001 Mean=0.001


Statistics on SWH

SWH(LSE-GDR) SWH(MAP-GDR)

Skewnesssolved

Skewnessfixed

Mean=7.64 cm Mean=7.91 cm

Mean=28.8 cmMean=7.06 cm


Jason-1 SLA Power Spectrum

J1 LSE5

TP LSE5

J1 MLE4 (GDRB)

J1 LSE4 (fixed skew)

Very good coherence between Topex LSE5 and Jason MLE4 Between MLE4 and LSE4, impact of 1HZ estimation for SWH, Pu and Att2 (skew fixed for both) Between LSE4 and LSE5, impact of the estimation of a 5th parameter (s) Between MLE4 and LSE5, addition of the two previous effects (it has been shown that they are very close regarding the SSB)

(see Faugere’s talk)


Conclusions

• LSE and MLE4 algorithms are equivalent wrt SSB when skewness is solved for.

• SSB differences between Topex and Jason cannot be explained by differences in the retracking algorithms

• SSB differences lie in the WFs themselves • SWH biases (LSE) are still to be analysed • MAP is not yet ready to provide users with reliable estimates

Range_LSE-Range_GDR versus (SWH,SIG0)

Skew solved

Simulator of Interferometric Radar Altimeters

concept and first resultsP. THIBAUT, B.PICARD : CLS, France, O.GERMAIN : Starlab, Spain, F.COLLARD : Boost-Technologies, France

L.PHALIPPOU : Alcatel-Alenia-Space, France C.BUCK5 : ESTEC-ESA, The Netherlands

PRF-1

Vsat

Left swathRight swath

Nadir track

Sea State Modelling

Waveform generation

Boost

Starlab

InterferometricInversion

CLS

CLS


Sigma-0 blooms in the Envisat Radar Altimeter dataP.THIBAUT, F.FERREIRA : CLS, France, P.FEMENIAS :

ESA/ESRIN, Italy

Which one is a bloom, which one is not ?Where are the egdes ?What are their length ? Where are they ?What about the waveforms during blooms ?What are the impact of blooms on ranges ?How many blooms are edited by Calval criteria?Which kind of criteria would be more useful ? ….Ideas, comments …. Come to talk with me.

evaluation of the ground retracking algorithms on jason-1 data

Documents

swh smaller

function of swh skewness

data skewness

swhlsemap skewness

att skew solved skewness

gdr ranges

att skewness solvedmean

epochjpldelta rangelse