new post-processing methodology for altimetry in coastal...

NEW POST-PROCESSING METHODOLOGY FOR ALTIMETRY IN COASTAL AREAS

Laurent ROBLOU (LEGOS)In collaboration with:

Jérôme BOUFFARD, Stéfano VIGNUDELLI, Florent LYARD, Matthieu Le Hénaff, Julien LAMOUROUX

New processing strategies for altimetry in coastal regions Roblou et al. 06/02/2008 2

Satellite altimetry in the coastal domain

Is satellite altimetry possible in the coastal domain?

ProblemsIncreased space/time sub-sampling:

Shorter horizontal scalesAmplified high frequencies

Increased error budgetIncreased loss of data

SolutionsNew sensorsMulti-mission approachImprove data post-processing


Data post-processing: example of the X-TRACK processor

The objective of the X-TRACK processor is to improve both the quantity and quality of altimeter sea surface measurements in coastal regions:

Features:redefining the data editing strategy to minimize the loss of data during the correction phase using improved local modelling of tidal and short-period atmospheric forcingUsing an accurate mean sea level consistent with the coastaldata set


X-TRACK historical playground: NW Mediterranean Sea


X-TRACK features: a new data editing strategy (1)

The X-TRACK processor adopts a new data screening strategy and filtering techniques allowing to recover data that would otherwise be flagged as badDe-flagging and re-interpolation of each single correction yields a reconstructed level profile.

Circles: uncorrected sea level anomalies (SLA) and original corrections from the AVISO Geophysical Data Records (GDR).

Brown line: SLA after application of the standard corrections from the GDR.

Purple line: the new SLA profile computed with X-TRACK processor

Roblou (2004)



number of T/P cycles exploitable

X-TRACK SLA productAVISO DT-(M) SLA product

Cal/Val sitesBirol (pers. comm.)



More altimeter data near to and far from the coastCoherent behaviour for ascending/descending passesTime series are longer (+10%) and less noisy (-7%)

Bouffard et al.

Gulf of Lion Gulf of Lion

X-TRACKSLA

AVISODT-(M) SLA

Lack of coastal data

95 120 95 120

North

44°

43°

42°

North

3° 4° 5° 3° 4° 5°


X-TRACK features: regional de-aliasing corrections (1)

Issues:Aliasing of the tides and short-period ocean response to meteorological forcing is a major problem when estimating the seasonal or longer time scales oceanic circulations in altimeterdataBecause of their insufficient spatial resolution that implies unresolved rapid changes in tidal features and an incorrect frictional dissipation, current global model cannot represent tides over continental shelves below a decimetre error level.

Solution: defining regional modelling of tides and ocean response to atmospheric forcing.

T-UGO 2D model (Mog2D model follow-on) for both processes



LEGOS regional models

Caspian Sea

Amazonian shelf



M2 mean error: 1.3 cm rms (TG), 0.8 cm rms (ALT)K1 mean error: 1.0 cm rms (TG), 0.9 cm rms (ALT)

Roblou (2003)

Mog2D tidal regional models

M2 mean error: 28.4 cm (TG), 5.3 cm (ALT)K1 mean error: 1.7 cm (TG), 1.2 cm (ALT)M4 mean error: 5.7 cm (TG), 1.2 cm (ALT)

Letellier (2004),



Aliased (<20 days) signal reduction estimates

0

10

20

30

40

50

60

70

80

90

barce

lona

ceuta

chios

corsi

nidu

brovn

ikha

dera

katak

olomala

gamars

eille

monac

onic

epa

lma

rhode

sse

netos

ase

te split

toulon

trieste

valen

ciaza

dar

kope

r

tide gauge

CM

**2

detided sea level variance (cm**2)

idem with IB correction

idem with model correction

Comparisons to tide gauges data:Gain vs IB: 46%Gain vs Mog2D-G correction: 5%

Similar results w.r.t altimetry (Bouffard)

Lyard and Roblou (2004)

Response to wind and atmosphericpressure forcing

Mog2D model


X-TRACK features: an accurate mean sea level (1)

Until accurate estimates of the geoid small- and meso-scales undulations become available, the dynamic topography of the ocean is not fully accessible in the altimeter measurements. Thus, the ocean mean sea surface provides an alternative reference surface, suitable for the observation of the ocean variability. Standard MSS products are not accurate in coastal regions

Solutions:Higher resolution MSS (including across-track effects) computed from an inverse methodMSS consistent with the improved altimeter products


X-TRACK features: an accurate mean sea level (2)

Sea level anomalies versus latitude

SLA=SSH-MSS CLS01

SLA=SSH-MSS X-TRACK

Differencecoast

MSS CLS01

Jaso

n-1

pass

009

MSS X-TRACK

MSS products comparisons


X-TRACK work in progress : orbit error reduction

Time evolution of the mean along-track SLA, GFO 074

Step 1: Raw signal vs Low Frequency (LF) signal,

Step 2: Localization and elimination of the outliers: 3 σ filtering of the High frequency signal,

Step 3: Re-building of the total corrected signal: linear regression.

Bouffard et al


X-TRACK work in progress: high rate data stream

Averaged along-track spatial spectrum, T/P track 146.

Bouffard et al


Jason-1 sea surface height calibration experiment

Jason-1 pass 222

Jason-1 pass085

M3 tidegauge

(Senetosa)

SSH absolute bias at M3 tidegauge, Jason-1 pass 085

0

0,02

0,04

0,06

0,08

0,1

0,12

0,14

0,16

0,18

0 20 40 60 80 100 120 140 160

Cycle

SSH

abs

olut

e bi

as (m

)

Mean=0.1046m

Std dev=0,0252m

Official estimates (Jason-1 CalVal teams): 0.1019 +/- 0.0268 m


Tidal analysis of X-TRACK SLA products vs FES2004 solutions

Tidal constants data set derived from T/P X-TRACK SLA colocated onto mean tracks (by harmonic analysis)

K1 misfit: 0.7 cm rmsM2 misfit: 0.7 cm rms

Misfit spatially constant over the whole domain, no differencies betweencoastal and deep ocean

Very few hot spots (mainly due to land contamination in the altimetermeasurement )

Tidal misfits:

background chart amplitude in cm from FES 2004 model.

The size of the red circles is proportional to the modulus of the complex difference between model and observation


GFOPass 31

JASONPass 146

T/PPass 009

ENVISATPass 0302

Good agreement between the SYMPHONIE coastal model (red curve) and the X-TRACK improved altimetry data (blue curve) in the typical synoptic scale and meso-scale.

Model seems less efficient in reproducing shorter space scales

SYMPHONIE model vs X-TRACK altimeter data in the NW Med (1)


Jason-1Pass 146 Time-averaged spatial along track spectra

- Good agreement between X-TRACK altimeter data and the Symphonie model on spatial scale greater than 50 km. Statistical agreement for shorter scales.- Very small spatial scales (<13 km) are not reproduced by the SYMPHONIE model

February May

August November

Seasonal along-trackcomparisons

Model unsolveddynamics

Model

Altimetry

Model

Altimetry

2/ SYMPHONIE model vs X-TRACK altimeter data in the NW Med (2)


SYMPHONIE model vs X-TRACK altimeter data in the Bay of Biscay

Blue: altimeter data

Red: Symphonie coastal model

Sea level anomalies comparisons in the Bay of Biscay along Jason-1 ground track 137 (July-August 2004).

good agreements on instantaneous sea levels

satisfying correlation for synoptic scales and meso-scales dynamics

altimeter data exhibit short scales processes not represented in the model simulations

This promising result opens up the way to an operational monitoring of coastal environnement (GMES framework), provided an overall improvement of short scales representation in coastal models.


Summary

New post-processing strategy for improving altimeter data for applications in coastal and shelves seas

Along-track editing (instead single point approach)High rate data streamRegional de-aliasing for astronomic and atmospheric loadingImproved vertical reference frameLarge scale errors reduction

Validated along track products for some regional seasNext step

Investigate other places/dynamicsTest retracked products (pre-processing as an input)Data assimilation into coastal models

new post-processing methodology for altimetry in coastal...

Documents