ifremer planning of cal/val activities during in orbit commisioning phase

Ifremer Planning of Cal/Val Activities during In orbit commisioning Phase

N. Reul, J. Tenerelli, S. Brachet, F. Paul & F. Gaillard, ESL & GLOSCAL teams

In orbit commissioning phase meeting ESAC 10/2009

There is no history of L-band Tbs over the global ocean

Previous & currently orbiting EOS low MW frequency radiometers (TMI,AMSR,WindSat) are (1)classical radiometers and (2) Operating at higher frequencies (C, X bands)=>low noise Tbs data (0.1-0.5 K)But very low sensibility to SSS (10-40 times less sensible than at L-band)

SMOS is : an interferometric radiometer and the reconstruction process accuracy is key for the

SSS retrieval success much more sensible to SSS than previous sensors BUT 20 times more noisy than previous sensors (2-5 K)

At the beginning of the mission, although they might be in errors, when can onlyRely on: Predictions from forward emissivity models applied to a priori « best

known » geophysical conditions viewed by the instrument, AND to account for instrumental imaging features

Context:

Main activities during commisioning:

=>First Priority activityValidate « calibrated » L1c Tbs & reconstruction process accuracy over the ocean based on forward model estimates:Focuss on sunglint & land effects

=>Second priority:Conduct Level 2 validation by comparison with in situ& reference fields

How to reach our first priority aims?

=>First: look @ averaged measured dependencies (incidence, polarization,SSS, SST, wind speed) and compare them to the expected ones from

forward models. Separation by pass type (asc,desc) & polarization mode

Ideally, what we would like to do:SMOS data

Globally averagedExpected meanFrom forward modelFor the period & Spatial domainconsidered

In practice, however:

1) we cannot do that in the surface frame but only in the antenna frame (because of polarization mixing and singularities when going from antenna to surface).

AND

2) We have to take into account and monitor the image Reconstruction process impacts on the Tbs

=> This definitively requires an end-to-end simulation strategy

Geophys Aux Data(SMOS ECMWF or external data

If not yet available)

L1C Aux Data

Sets of SMOS L1C Data

Forward Model Package

Modeled Tx,Ty,Uxy,Vxywithout reconstruction

Modeled Calibrated Visibilities

Modeled T’x,T’y,U’xy,V’xy

including reconstruction effects

Image Reconstruction soft

TRAP SMOSSoftware

TRAP-SMOS

•The software is an extension of the aircraft analysis package, TRAP, that will enable some basic analysis of SMOS data down to level 1a.

•Based on a set of C++ classes representing the calibrated visibilities, the G matrix, the correlations, and the J+ matrix.

•Currently a basic MEX interface provides a means of computing calibrated visibilities and then inverting them to obtain a reconstructed scene for an arbitrary scene provided by the user. This can be done zith just a few lines of matlab code.

•For the G matrix, unix memory mapping is used to enhance efficiency, allowing the entire G matrix to reside in memory even on a 32 bit machine (with sufficient memory). As only one sub-block of G is mapped into the process at any one time, 64 bit addressability is not needed. Also, multiple processes can share the same G matrix.

TRAP-SMOS: Two Examples of what we will do

•Sun glint contamination: the current L1 Processor uses a fixed wind speed to compute glint contamination. Is this a serious source of error?

•Land contamination: how might the spatially varying (in each snapshot) bias evolve with changing land distribution?

Possible Sunglint Contamination

Kirchhoff scattering model; 7 m/s wind speed

Possible Sunglint Contamination

Kirchhoff scattering model; 20 m/s wind speed

Possible Sunglint ContaminationInstrument Reconstruction: Kirchhoff scattering model;

7 m/s wind speed

Possible Sunglint Contamination

Instrument Reconstruction error: 7 m/s wind speed

Possible Sunglint Contamination

Instrument Reconstruction error: 20 m/s wind speed

Land Contamination

Possible Land Contamination

Possible Land Contamination

Decomposition of Reconstruction

Decomposition of Reconstruction

Here much of the spatially varying error is

associated with brightness beyond the fundamental hexagon.

IFREMER OS ESL: Plan of Action up to KP1

•Once we obtain calibrated visibilities after launch we can begin to compare modeled and instrument reconstructed scene brightness in a variety of circumstances. For example, over uniform sea do we observe a consistent bias pattern consistent with that implied by the measured antenna patterns? Can we detect the distant impact of land in the measurements?

•It may be very difficult to extract the important biases from the noise and we are still considering how to achieve this.

•We would like to be able to interact with Deimos and the Level 1 ESLs as we examine the visibilities. We do not expect to address issues below level 1a and therefore such interaction will be important if we cannot reconcile the forward models and instrument at a higher level.

•=> Joe tenerelli will visit the C-EC at least up to KP1-KP2

IFREMER OS ESL:

Plan of Action after KP1:

J. Tenerelli=> Continue on previous ocean scene bias activities

N. Reul and S.Brachet J. Tenerelli:

=>Look @ local and expected « strong » or « weakly variable »local patterns in the Tropical ocean

=>Starts L2 validation if possible, with in situ in these zones

In practice, we will try to implement this analysis in order to assess

Our ability to detect expected Large scale Hot & Cold spots over the oceanOver monthly periods

We will need to perform Tb space-time averaging to detect stable geophysical & noise-generated features on local scales

Time Averaging

Expected monthly averaged surface Tbs @ L-band, nadir andfor January

Hot Spots

Cold Spots Expected Range

To study Local Tbs behaviour:

Analyze simplest conditions by selecting data for which :

- we will try to minimize anticipated dominant image reconstruction problems:

Land

Sunglint

-we will try to avoid strong geophysical contaminations and difficult situations:

Galactic & sunglints,

low & high winds,

low sst (tropics),

rain

Other Local « strong source » validation site:

The dipole of the Northern Indian Ocean

Constitution of a validation database including:*Available SMOS L1C & L2

*Co-located geophysical products:

– SMOS ECMWF

– climatologies of SSS & AMSR-E SSS over the tropics

– Satellite Sea Surface Temperature

– Satellite & blended winds,

– Wave model products, …

• +Qualified in situ data

• =>Match-up Data Base analysis of

L1C, L2 & L3

Built up of a validation database

Control L2 large-scale behaviour with reference climatological fields over identified ocean basins

Match-up with In situ: accounting for the natural Variability

In high variability zone: in situ need to be compared to 1 day to 10 days-averaged SMOS data

Data that will fill the match-up Data baseAnd L2 Validation processFor the commissioning

Statisitics from the match-up per in situ sensor type: SMOS versus Argo profilers

Statisitics from the match-up per in situ sensor type: AMSR versus moorings

Statisitics from the match-up per in situ sensor type: SMOS versus Piratta moorings

Merged Statisitics from all the match-up : SMOS versus all in situ data

Merged Statisitics from all the match-up : SMOS versus all in situ data:Rms error per 1 psu SSS bins