Smos ocean salinity retrieval level 2
This presentation is the property of its rightful owner.
Sponsored Links
1 / 57

SMOS Ocean Salinity Retrieval Level 2 PowerPoint PPT Presentation


  • 90 Views
  • Uploaded on
  • Presentation posted in: General

SMOS Ocean Salinity Retrieval Level 2. Marco Talone, Jérôme Gourrion, Fernando Martin Porqueras, Nicolas Reul, Joe Tenerelli, Marcos Portabella, and the SMOS Barcelona Expert Centre (SMOS-BEC) Team [email protected]

Download Presentation

SMOS Ocean Salinity Retrieval Level 2

An Image/Link below is provided (as is) to download presentation

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.


- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -

Presentation Transcript


Smos ocean salinity retrieval level 2

SMOS Ocean Salinity RetrievalLevel 2

Marco Talone, Jérôme Gourrion, Fernando Martin Porqueras, Nicolas Reul, Joe Tenerelli, Marcos Portabella, and the SMOS Barcelona Expert Centre (SMOS-BEC) Team

[email protected]

Course on Earth Observation Understanding of the Water CycleFortaleza, Brasil, November 1-12, 2010


Smos ocean salinity retrieval level 2

Motivation


Ocean salinity monitoring motivation overview

Ocean salinity monitoring: motivation/overview

  • SSS variations governed by:

    • E-P balance

    • freezing/melting ice

    • freshwater run-off

  • Key oceanographic parameter (density)

  • Thermohalinecirculation and heat redistribution


Ocean salinity monitoring motivation overview1

Ocean salinity monitoring: motivation/overview

Historical lack of SSS observations

10-m depth salinity field reconstructed from Argo floats data. There are still “holes” and spatial resolution is low

SSS time-series

Surface salinity distribution is closely tied to E-P patterns


Ocean salinity monitoring motivation overview2

Ocean salinity monitoring: motivation/overview

  • Oceanographic models already assimilate SST and SSH from satellite data, while SSS is still climatologic

  • The absence of any specific treatment of salinity in ocean models can lead to significant errors:

    • Near-surface currents errors [Acero-Schetzer et al., 1997]

    • Tropical dynamics [Murtugudde and Busalacchi, 1998]

    • Dynamic height difference [Maes et al., 1999; Ji et al., 2000]

    • Spurious convection [Troccoli et al., 2000]

    • ENSO predictions [Ballabrera-Poy et al., 2002]


Smos ocean salinity retrieval level 2

SMOS, general features


Smos satellite general features

SMOS satellite – general features

  • 1.4 GHz, L-band (unique payload)

  • Optimum SSS sensitivity

  • Reasonable pixel dimension

  • Atmosphere almost transparent

  • Synthetic Aperture Radiometer (MIRAS)

  • Sun-synchronous LEO orbit, 3 days revisit time

  • 69 elements array, Y-array: arms 120º apart

  • Field Of View (EAF FOV) about 1000 km

  • Dual-pol / Full-pol

  • Multi-angular capabilities

  • Spatial Resolution: 32 (boresight) - 100 km

  • Full scene acquired every 2.4 s

  • Variable number of observations according to the satellite sub-track distance

  • Different measurements of TB corresponding to a single SSS under different incidence angles


Smos satellite field of view

SMOS satellite – Field Of View

boresight

nadir

Radiometric Accuracy and Radiometric Sensitivity (quality of the measurement)

[calculated using SEPS]

Incidence Angle and Spatial Resolution

[calculated using SEPS]


Smos satellite field of view1

SMOS satellite – Field Of View

  • Due to MIRAS geometry Nyquist criterion is not satisfied

  • 3 FOV can be defined:

  • Hexagon resolved by MIRAS

  • Alias-Free FOV

  • Extended Alias-Free FOV


Smos ocean salinity retrieval level 2

SMOS processing chain


Smos processing chain

SMOS processing chain

Raw data

Measurements

Observations

Global map

Data fusion

Level 0

Level 1

Level 2

Level 3

Level 4

Data Assimilation


Smos processing chain1

SMOS processing chain

Level 0 Raw data

Level 1A Calibrated Visibilities

Level 1B TB Fourier components

Level 1C TB geocoded (ISEA4H9)

Level 2 Salinity Maps (single-overpass)

Level 3 Spatio-temporal averaged SSS

Level 4 Merged product

  • Scientific requirements for salinity retrieval

  • Global Ocean Data Assimilation Experiment (GODAE, 1997)

    • 0.1 psu, 200 km, 10 days

  • Salinity and Sea Ice Working Group (SSIWG, 2000)

  • 0.1 psu, 100 km, 30 days

  • SMOS (Mission Requirements Document v5, 2002)

  • 0.1 psu, 200 km, 30 days

  • lower accuracy, higher resolution products (e.g. 100 km, 10 days or single passes) are useful for applications other than climate and large scale studies

ISEA DGGs (Discrete Global Grids)


From level 1c to level 3

From Level 1C to Level 3

Level 1C

Level 1C

pre-

processing

Level 2

post-

processing

Level 3

quality control

& filtering

SSS

inversion

Level 1C


Level 1c

Level 1C

What does a radiometer measure?

Boltzmann constant

G accounts for the receiver’s gain

and the antenna pattern

receiver temperature

is the only term dependent on the observed scene

it is also referred as Apparent Temperature because sum of various contributions


Level 1c forward models

Level 1C - Forward Models

ionosphere

atmosphere

surface


Level 1c forward models1

Level 1C - Forward Models

3 MODELS

Two-scale model IFREMER Brest, France

Small Slope Approximation (SSA) model LOCEAN, Paris, France

Empirical Model ICM, Barcelona, Spain

roughness

contribution

flat sea contribution

Klein & Swift (1977) dielectric model at microwave frequencies

The total dynamic of TB is 2-4 K

TB sensitivity to SSS increases with SST


Level 1c forward models2

Level 1C - Forward Models

Brightness temperature as measured by SMOS

TY

TX


From level 1c to level 31

From Level 1C to Level 3

Level 1C

Level 1C

pre-

processing

pre-

processing

Level 2

post-

processing

Level 3

quality control

& filtering

SSS

inversion

Level 2 pre-processing


Level 1c errors and inaccuracies

Level 1C – Errors and inaccuracies

Several different phenomena contribute to the final

The main error sources for the SSS retrieval are:

  • The forward Tb models

  • The estimation of the antenna pattern

  • The estimation of the galactic noise

  • Radio Frequency Interference

  • Land contamination

Some of them are solved by pre- and post-processing techniques


Level 2 pre processing

Level 2 pre-processing

Ocean Target Transformation

Average instrumental spatial pattern against ocean target, to be subtracted from

measurements prior to SSS retrieval.

[J. Tenerelli, Tech Note, 2010]

INCLUDED IN THE CURRENT

PROCESSING

  • An accurate filtering of the snapshots must be applied to discard land and/or Radio Frequency Interferences (RFI) contaminations.

  • Ascending and descending passes must be considered separately.

  • Finally, many orbits are used to increase the robustness of the estimation.


Level 2 pre processing1

Level 2 pre-processing

Strong systematic patterns are found in SMOS TB measurements

Features are clearly associated to brightness temperature transition:

Sky/Land

Alias Free/Extended Alias Free Field of View


Level 2 pre processing2

Level 2 pre-processing

The use of a forward model can introduce error due to inaccuracies in its definition

Unhomogeneities in the geophysical parameter statistical distribution in the FOV affect the estimation of the OTT

INCLUDED IN THE NEXT

REPROCESSING (july)

Model-free OTT – X pol

Model-free OTT – Y pol


Level 2 pre processing3

Level 2 pre-processing

Sea Surface Temperature

[°C]

Wind Speed

[m/s]

Histograms are calculated for all the pixel of the reconstructed brightness temperature image (black lines).

A selection of the grid point used in the averaging is performed to homogenize all the histograms the most internal one (red line)


Level 2 pre processing4

Level 2 pre-processing

STANDARD

HOMOGENIZED

The average sea surface salinity, sea surface temperature, and wind speed inside the FOV are shown for the standard OTT and the “homogenized” OTT.


Level 2 pre processing5

Level 2 pre-processing

The difference between the “homogenized” and “no-homogenized” OTT.

Model-free OTT – X pol

Standard OTT – X pol

MODE

L

-FREE

STANDARD

Model-free OTT – Y pol

Standard OTT – Y pol

1 - 2 °C for sea surface temperature

and 0.5 - 1 m/s for wind speed

up to 0.6 - 0.8 K (peak to peak) in the estimation of the bias spatial pattern.


Level 2 pre processing6

Level 2 pre-processing

External Brightness Temperature Calibration

Average instrumental temporal pattern (scene-dependent bias) against ocean target, to be subtracted from measurements prior to SSS retrieval.

[Camps et al., Radio Science 2005]

IN TESTING PHASE


Level 2 pre processing7

Level 2 pre-processing

  • Radio Frequency Interference (RFI)

  • Spurious stable or intermittent man-made interferences.

  • Receiver Co-Channel Interference + Receiver Adjacent Signal Interference - the signal itself or its tails can fall within the receiver’s RF passband.

  • Receiver Out of Band Interference - the signal is outside the receiver’s RF passband, nevertheless spurious signals due to the mixer stage.

  • Transmitter Fundamental and Harmonic Emissions - the Transmitter Transfer Function.

  • Transmitter Noise - thermal noise generated in the various stages of the processing.

  • Transmitter Intermodulation - local mixing of a transmitter’s output emission with that of another transmitter or any other component of the instrument.

  • Concerning SMOS the strongest interference come from WiFi networks and Radar

  • As expressed in the Technical Note on “L-band RFI detected in SMOS data over the world oceans” by Nicolas Reul of IFREMER.


Level 2 pre processing8

Level 2 pre-processing

By estimating the impulsional response of the RFI, this can be eliminated from the scene,

as done for the Sun effects.

[Camps, 2010]

INCLUDED IN THE NEXT

REPROCESSING (july)


From level 1c to level 32

From Level 1C to Level 3

Level 1C

Level 1C

pre-

processing

pre-

processing

Level 2

Level 2

post-

processing

Level 3

quality control

& filtering

quality control

& filtering

SSS

inversion

SSS

inversion

Level 2


Quality control and filtering

Quality control and filtering

Quality control is performed on both measurement and gridpoint basis

  • Distance from the coast:

  • Ice

  • Suspect ice

  • Heavy rain

  • Sea condition

  • Number of valid measurements

  • Sunglint

  • Moonglint

  • Galactic noise

  • position in the FOV:

  • RFI

Land

< 40 km

40 km - 200 km

> 200 km

OK

Retrieved but Flagged

Discarded

AF

EAF

Border FOV

Aliased FOV


Sss inversion

SSS Inversion

The problem

state variables

observations

forward model

The solution

  • exact algebraic solution,

  • relaxation,

  • least squares estimation,

  • truncated Eigenvalue expansion,

  • Bayes’ theorem,

  • etc …

  • maximum likelihood,

  • maximum posteriori probability,

  • minimum variance,

  • minimum measurement error

  • etc …


Sss inversion theoretical background

SSS Inversion - theoretical background

Bayesian approach

posterior probability

uncertainty of observations

andforward model

knowledge about the

state variables

ASSUMING AND INDEPENDENT (ERRORS UNCORRELATED)


Sss inversion theoretical background1

SSS Inversion - theoretical background

Maximum Likelihood Estimation

Errors are generally assumed Gaussian

1.Forward Model (GMF, Geophysical Model Function)

is assumed perfect

2.Errors are assumed uncorrelated

is diagonal


Smos sss inversion

SMOS SSS Inversion

SMOS Sea Surface Salinity Retrieval Cost Function

observables part

Background term

is minimized iteratively

YES

min

NO

INITIALIZATION


From level 1c to level 33

From Level 1C to Level 3

Level 1C

Level 1C

pre-

processing

pre-

processing

Level 2

Level 2

post-

processing

post-

processing

Level 3

Level 3

quality control

& filtering

quality control

& filtering

SSS

inversion

SSS

inversion

Level 2 post-processing


Level 2 post processing

Level 2 post-processing

External Sea Surface Salinity Calibration

Correcting for the mean uncertainty introduced by the forward model inaccuracies as done for rain radar calibration (Seo and Breidenbach, 2002) using as ancillary in-situ database the ARGO array of buoys.

[Talone et al., IEEE TGARS 2008]

IN TESTING PHASE


Smos ocean salinity retrieval level 2

Level 1C problems @Level2

RFI

Galactic Noise

Land contamination


Level 1c problems @level2 rfi

Level 1C problems @Level2 - RFI

North Pole case: Radar

Average SSS in April – ASCENDING PASSES

Average SSS in April – DESCENDING PASSES

SOURCE

anti-missile radar protection array from Alaska all along the Northern Canada pointing to the horizon

climatological SSS


Level 1c problems @level2 rfi1

Level 1C problems @Level2 - RFI

Due to the signal processing in SMOS, a point strong source generates 60-degrees spaced tails, like a star.

First Stokes’ parameter in brightness temperature (I=TX+TY). One-month averaging, only descending passes.

Retrieved SSS


Level 1c problems @level2 rfi2

Level 1C problems @Level2 - RFI

Due to the signal processing in SMOS, a punctual strong source generates 60-degrees spaced tails, like a star.

First Stokes’ parameter in brightness temperature, both the ascending and descending passes of the month of May


Level 1c problems @level2 rfi3

Level 1C problems @Level2 - RFI

RFI has effects several kilometers from the source. Sources on land frequently affects SSS retrieval.

First Stokes’ parameter in brightness temperature, both the ascending and descending passes of the month of May


Galactic noise

Galactic Noise

Scattering model for ocean surface reflection of downwelling celestial radiations

Very geographic, pass-type & incidence angle dependent

[Nicolas Reul, IFREMER, 2010]


Level 1c problems @level2 land cont

Level 1C problems @Level2 – Land cont.

The image reconstruction algorithm in SMOS is almost a FFT. Any sharp transition introduce singularities and its inversion introduce errors. Land’s brightness temperature is 300 K, while the average sea surface brightness temperature is 120 K.


Smos ocean salinity retrieval level 2

Level 2 Products


Level 2 products

Level 2 Products

SMOS Level 2 User Data Product – UDP is available, one file per semi-orbit, on:

http://eopi.esa.int/esa/esa (a data request form must be filled first)

proc

version

end

YYYYMMDDThhmmss

start

YYYYMMDDThhmmss

SM_OPER_MIR_OSUDP2_20101023T140558_20101023T145957_316_001_1.zip

SM_OPER_MIR_OSUDP2_20101023T140558_20101023T145957_316_001_1.HDR

header in XML

SM_OPER_MIR_OSUDP2_20101023T140558_20101023T145957_316_001_1.DBL

binary data file


Level 2 products1

Level 2 Products

Different programs are available to open, display, partially process, and export SMOS Level 2 data, among them:

BEAM software, including SMOS-box plug-in can be downloaded from

www.brockmann-consult.de

Binary .DBL files can be read by using ad-hoc programs (C, Matlab, Fortran…), exported data can feed any program you are most used to (IDL, Matlab, ODV…)

Details on DBL file structure can be found in the L2 Product Specification Document, on:

www.smos-bec.csic.es


Level 2 product

Level 2 Product


Level 2 product1

Level 2 Product

  • 3 retrieved SSS

  • 3 theoretical uncertainties associated to the 3 retrievals

  • Acard

  • theoretical uncertainty associated to the retrieval of Acard

  • Wind Speed

  • theoretical uncertainty associated to the retrieval of WS

  • Sea Surface Temperature

  • theoretical uncertainty associated to the retrieval of SST

  • Modeled Brightness Temperature at 42.5° pol H (surface)

  • theoretical uncertainty associated to TBH

  • Modeled Brightness Temperature at 42.5° pol V (surface)

  • theoretical uncertainty associated to TBV

  • Modeled Brightness Temperature at 42.5° pol X (antenna)

  • theoretical uncertainty associated to TBX

  • Modeled Brightness Temperature at 42.5° pol Y (antenna)

  • theoretical uncertainty associated to TBY

  • Control Flag: Several quality flags one for retrieval (4)

  • Dg_chi2: Retrieval fit quality index, one for retrieval (4)


Level 2 product2

Level 2 Product

  • Dg_chi2_P: chi2 high value acceptability probability,

  • one for retrieval (4)

  • Dg_quality: Descriptor of SSS uncertainty, one for

  • retrieval (4)

  • Dg_num_iter: number of iterations until convergence,

  • one for retrieval (4)

  • Dg_num_meas_L1c: number of measurements at L1c

  • Dg_num_meas_valid: number of valid measurements

  • after discrimination

  • Dg_border_fov: number of grid-points at the border of the

  • FOV

  • Dg_eaf_fov: number of grid-points in the Extended

  • Alias-Free FOV

  • Dg_af_fov: number of grid-points in the Alias-Free FOV

  • Dg_sun_tails: number of grid-points affected by sun

  • reflection tails

  • Dg_sunglint_area: number of grid-points affected by sun

  • reflection

  • Dg_sunglint_fov: number of grid-points with reflected sun

  • in the FOV


Level 2 product3

Level 2 Product

  • Dg_sunglint_L2: number of grid-points with reflected sun

  • in the FOV, as computed at L2

  • Dg_suspect_ice: number of grid-points with suspected ice

  • Dg_galactic_Noise_Error: number of grid-points affected by

  • galactic noise

  • Dg_galactic_Noise_Pol: number of grid-points affected by

  • polarized galactic noise

  • Dg_moonlight: number of grid-points with reflected

  • moonlight in the FOV

  • Science_Flags: several geophysical flags

  • Dg_sky: number of gridpoints with specular direction

  • toward a strong galactic source

  • Land_Sea_Mask: Land/Sea descriptor


Smos land cover tool

SMOS Land Cover Tool

Tool from GMV to display and export SMOS product to Google Earth files


Level 2 product4

Level 2 Product

Exercise


Thank you

Thank you!

Course on Earth Observation Understanding of the Water CycleFortaleza, Brasil, November 1-12, 2010

Marco Talone, Jérôme Gourrion, Fernando Martin Porqueras, Nicolas Reul, Joe Tenerelli, Marcos Portabella, and the SMOS Barcelona Expert Centre (SMOS-BEC) Team

[email protected]


Interferometric radiometer

Interferometric Radiometer

  • The idea at the basis of interferometric radiometry is to synthesize a large antenna with a number of small ones. Output voltages of a pair of antennas (e.g. located at and ) is cross-correlated to obtain the so-called “visibility sample”:

where ,

B1 and B2 are the receivers' noise bandwidths, G1 and G2 the available power gains, and b1(t) and

b2(t) the signals measured by elements 1 and 2.

The complete set of the visibility samples is called a visibility map, and it is approximately the Fourier transform of the brightness temperature distribution of the scene. To invert this process the inverse Fourier transform can be applied as a first approximation (Camps et al., 1997) or a more sophisticated G-matrix inversion (Anterrieu and Camps (2008); Camps et al. (2008a))

The result is a potential degradation of the radiometric sensitivity in terms of a higher rms noise, on the other

hand a complete image is acquired in one snapshot, permitting to increase the integration time and improve the measurement quality. Nevertheless, the major advantage of interferometric radiometry is the multi-angular measurement: the output of an IR is, in fact, an image; this permits having several views under dierent incidence angles of the same point on the Earth before it exits from the Field of View


Smos ocean salinity retrieval level 2

RFI

  • FROM http://www.radioing.com/eengineer/rfi.html

  • 1.0 Receiver Co-Channel Interference

  • This is defined as undesired signals with frequency components that fall within the receiver’s RF passband and are translated into the Intermediate Frequency (IF) passband via the mixer stage. The interfering signal frequency is equal to the sum of the receiver’s tuned frequency and one half of the narrowest IF bandwidth. These signals are amplified and detected through the same process as the desired signals; therefore, a receiver is very susceptible to these emissions even at lower levels.

  • Results: Receiver desensitization, signal masking, distortion.

  • 2.0 Receiver Adjacent Signal Interference

  • This is defined as undesired signals with frequency components which fall within or near the receiver’s RF passband and are translated outside of the IF passband via the mixer stage. These signals must be of sufficient amplitude to produce non-linear effects within the receiver’s RF amplifier or mixer stages. Some of the resulting non-linear response signals may be converted to the IF passband frequency via the mixer stage where they are amplified and detected through the same process as the desired signals. These become similar to co-channel interference signals at this point. The undesired emissions which are translated outside of the IF passband may still pass through the remaining receiver stages, if at high enough levels to survive the out-of-passband attenuation. They may then be processed by the detector. The predominant response for this case is desensitization.

  • Results: Non linear effects in the RF or mixer stages producing receiver desensitization, intermodulation and cross modulation.


Smos ocean salinity retrieval level 2

RFI

  • FROM http://www.radioing.com/eengineer/rfi.html

  • 3.0 Receiver Out of Band Interference

  • This is defined as undesired signals with frequency components that are significantly removed from the receiver’s RF passband. High level signals may produce spurious responses in the receiver if mixed with local oscillator (LO) harmonics to produce a signal falling within the IF passband. The spurious responses result from the mixing of an undesired signal with the receiver’s LO. The amplitude of these responses is directly proportional to the level of the undesired signals prior to mixing with the LO. The spurious responses in a receiver usually occur at specific frequencies. Any other out of band signals are attenuated by the IF selectivity.

  • Results: An undesired response created by the mixing of an undesired signal with the LO. The undesired signals which mix with the LO and are capable of being translated to the IF stages are the spurious response frequencies. These frequencies and their interference power levels are a function of the receiver’s susceptibility to these responses.

  • 4.0 Transmitter Fundamental Emissions

  • The transmitter’s fundamental output signal includes characteristics of the power distribution over a range of frequencies around the fundamental frequency. These are determined by the base-band modulation characteristics and are represented by a modulation envelope function. The primary parameter associated with the modulation envelope is the transmitter’s nominal bandwidth (3dB). This may be derived from the transmitter modulation characteristics (by Fourier analysis), measured, or from the manufacturer’s specifications. The power distribution in the modulation sidebands may be represented by a modulation envelope function showing the variation of power with frequency.


Smos ocean salinity retrieval level 2

RFI

  • FROM http://www.radioing.com/eengineer/rfi.html

  • 5.0 Transmitter Harmonic Emissions

  • The main concern with a transmitter’s harmonic emissions is the undesired signal outputs which are harmonically related to the fundamental signal rather than to other oscillator circuits. The relative power associated with the harmonic emissions may be modeled using data for the particular transmitter type. However, since harmonic output power can vary considerably from one transmitter to another for the same type and model, it should be represented statistically. Harmonic emission models may be derived from statistical summaries of measured data or from manufacturer’s equipment specifications. Transmitter spurious emission models for prediction of frequencies above the fundamental are based on harmonic emission levels. The modulation envelope must be represented for harmonics as was done for the fundamental.

  • 6.0 Transmitter Noise

  • Transmitter noise includes the output spectrum that is a result of the thermal noise generated in the driver and final amplifier stages as well as the synthesizer noise from lower level stages. This is a broad-band noise; however, it usually does not cover the immediate modulation sidebands. The level may be specified as the power per bandwidth as a function of frequency (dBm/Hz).

  • 7.0 Transmitter Intermodulation

  • These are the undesired signals that result from the local mixing of a transmitter’s output emission with that of another transmitter. The mixing usually occurs in the non-linear circuits of a transmitter whose antenna receives a high level of RF from another transmitter antenna in close proximity. The mixing products are radiated by the transmitter’s antenna as possible co-channel or adjacent signal interference signals


  • Login