SMOS mission: a new way for monitoring Sea Surface Salinity?

J. Boutin (1)

(1) Laboratoire d’Oceanographie et du Climat- Expérimentation et Applications Numériques (LOCEAN), PARIS, FRANCE

Thanks to T. Delcroix (LEGOS/IRD) and F. Petitcolin (ACRI-st)

How GOSUD and SAMOS data could help for SMOS Cal/Val?

(SMOS/GLOSCAL (Global Ocean sea surface Salinity :

CALibration and validation for SMOS) project)

(IFREMER, LEGOS/IRD, LOCEAN/SA/IPSL, Meteo-France, CLS, ACRI-st)

SMOS (Soil Moisture and Ocean Salinity)

• Should be launched in 2007

• Goal: SSS accuracy: 0.1-0.2psu over 200x200km2 10days

• L-band radiometer (=21cm) =>SSS of upper 1cm depth

• Synthetic Aperture radiometer => spatial resolution 40km

• 3 arms => bidimensional field of view

animation

S1 T0

S10 24s

S20 48s

S30 72s

S40 96s

S50 120s

S60 144s

S70 168s

S80 192s

S90 216s

SEPSv3simulations

SMOS 2-D FIELD OF VIEW (one over 10 FOV), (F. Petitcolin, Acri-st)

Satellite passes at 6AM and 6PM UTC

Sensitivity of L-band Tb to SSS

(flat sea)

0.5K/psu (15°C)

0.7K/psu (30°C)

0.2K/psu (0°C)

Flat sea (Klein and Swift model)

Sensitivity of Tb to SSS is:

-small: always less than 1K/psu(SMOS radiometric precision of

1 Tb: several K)

-Higher in warm water

NB: L-band radiometer measurements are representative

of top 1cm surface ocean

Brightness temperature of the sea surface for a rough sea surface

0.5K/psu (15°C)

0.7K/psu (30°C)

0.2K/psu (0°C)

Flat sea Rough sea (without foam)

~0.2K/m/s

2 scale emissivity model: small waves superimposed on large tilted waves

Dinnat et al., IJRS, 2002, Radio Science, 2003

At 15°C, a 0.1K Tb variation can be generated by :

-0.2psu SSS variation or

- 0.5m/s wind speed variation

10m equivalent neutral wind speed (m/s)

SSS retrieved from multiangular Tb measurements

Cost Function to be minimized:

Tbmod: Tb estimated with a direct forward modelN: number of Tb observationsP: geophysical parameters responsible for Tb variations (e.g.: SSS, SST, wind ; depend on

forward model)i: errors on Tbmeas

k: Prescribed errors on auxiliary parameters (typical values): U = 2m/s; SST = 1°C

Retrieved parameters: SSS, SST, equivalent neutral wind speed (depend on forward model)

Minimization: Levenberg-Marquardt algorithm

2

1

0

2

1

mod2

K

k k

kkN

i i

imeasi PPTbTb

An iterative retrieval algorithm is used to retrieve SSS, SST, surface roughness parameters the most consistent with Tb measurements

1 satellite pass - 40x40km pixels (1 to 3K random error on individual Tb, U error 2m/s, SST error 1°C)

0.5

1

1.5

2

2.5

3

0 100 200 300 400 500 6000

5

10

15

20

25

30

X (km)

SSS

(Tx,Ty,W,SST)

11.1

SSS Error

Error on retrieved SSS (estimated with Dinnat et al. Model)

Boutin et al., 2004

Distance across track (km)

SS

T (

°C)

Error on retrieved SSS averaged in ‘Godae’ boxes (200x200km2 ; 10 days )

0.15PSU

Encouraging simulation but ‘optimistic’ hypothesis: -random noise on Tb and auxiliary parameters

-knowledge of the true forward modelTo be checked during Cal/Val in 2007!!!!

Boutin et al, JAOT,2004

0.20.1

Number of retrieved SSS in GODAE box (200km x 200km over 10 days)

400 1000 0.04 0.3

Error on mean SSS (computed as 2/N)

1) Estimate SMOS SSS accuracy and precision:-Compare SMOS SSS with in-situ SSS Need for SSS data

2) Identify error and biases sources: -flaws in direct emissivity models / instrument drifts

-Compare SMOS Tb with Tb derived from direct forward models Data needed to compute Tb: SSS, wind, SST, atmospheric

pressure, Tair Other useful information: Rain, wave, swell, currents

-flaws in auxiliary parameters (coming from ECMWF model/Reynolds analysis) used in the SSS inversion

-Compare them with in situ data Need for wind,SST,Patm,Tair

Goals of CAL/VAL using in situ data

Sampling, Precision and Accuracy of in situ data well adapted for SMOS Cal/Val depends on: Sensitivity of SMOS retrieved SSS to biases on auxiliary parameters

Natural variability of SSS (and auxiliary parameters)

SSS bias as a function of wind speed bias

Simulations of SSS retrieved from biased wind speeds

SSS bias mostly related to wind speed bias (at 15°C, 1psu bias <-> 2m/s bias);

WIND SPEED BIAS (m/s)

In order to get SSS bias<0.1psu , need for bias on wind speed data < 0.2m/s

SSS

BIA

S (

m/s

)

-8

8

-10 10

SSS bias as a function of SST (SST bias=5°C)

Influence of SST bias on retrieved SSS

SSS bias strongly dependent on SST: almost no bias around 15°C; >0 biases at low SST and <0 biases at high SST

In order to get SSS bias < 0.1psu, need unbiased SST especially at low and high SST: -at SST=30°C: SST bias<0.5°C

-at SST=0°C: SST bias<0.3°C (extreme value!)

Est

ima

ted

– r

efer

enc

e S

SS

(ps

u)

-8

-6

-4

-2

02

4

6

8

10

2 7 17 12 22 27SST (°C)

10day-horizontal variability of SSS as detected by ARGO floats

At 10 day interval ARGO floats drift over 56km on average (up to 200km in frontal regions) => difference between SSS recorded at 10 day interval by the

same float represents SSS variability at 10 day-20km to 200km scale

0 20060 140

10day-horizontal variability of ARGO measurements

(SSS10days )=0.2psu; largest differences in tropical regions;

similar results for 10days-20-50km and 10days-50-200km drift =>Number of measurements needed to achieve an accuracy of (a=0.1psu) on a 10day-

20-200km mean: N = 4 2 / a2 => N = 16 observations

Quadratic mean of SSS10days in 2°x2° pixels(N>10) July 2004-July 2005

0 0.30.1 0.2

Difference in SSS measured by the same float at 10 days interval

0.4 0.2 -0.4 -0.2 0SSS10days

Boutin and Martin, 2006

0

Delcroix et al., 2005

SSS variability derived from ships and moorings measurements

Estimate spatial SSS variability from ship measurements and temporal SSS variability from mooring measurements

SMALL SCALE VARIABILITY IN THE TIME DOMAIN, 0-165E

0

Delcroix et al., 2005

SMALL SCALE VARIABILITY IN THE SPACE (N-S) DOMAIN (PX04; Fiji-Japan line)

SSS variability derived from ships and moorings measurements

=> The mean expected variability within a box of 1°x2°x10 days is 0.2 psu => Nmin to achieve 0.1psu accuracy: N = 4 2 / a2 => N = 16 observations

The mean standard deviation of SSS over :-1° latitude is 0.1 psu

-2° longitude is 0.12 psu-10 days is 0.10 psu

(such values are variable in space and time)

Summary: requirements on in situ measurements for SMOS/Cal Val

Sampling Accuracy Precision Remarks

Parameters needed to compute Tb with present forward models (resolution ~40km):

SSS Nmeas=16 in 40x40km pixel (optimal)

0.05psu 0.1psu Depth: Upper layer (same as SST)

SST Similar to SSS 0.3°C 0.5°C Depth: Upper layer (diurnal cycle) (L-band signal coming from 1cm depth)

Wind speed (direction)

hourly 0.2m/s 1m/s Computed at 10m height (equivalent neutral)

Additional information very useful for interpreting SSS differences:

Rain;

Surface roughness: Currents; Waves and swell

Conclusions/Remarks

Advantages of GOSUD/SAMOS measurements w/r to other measurements:

-w/r to moorings: almost ‘global ocean’coverage : sampling of very variable meteorological and oceanographic conditions

-w/r to ARGO floats: provide meteorological measurements and complementary information necessary for interpreting differences between in

situ and SMOS SSS.

Remarks: It would be very convenient to get colocated ocean surface and meteorological

parameters or software generating colocated measurements.

Colocations useful for other applications? -Study of air-sea interactions (e.g. CO2 air-sea flux in case ocean CO2

measurements, see Lefevre et al. poster)…

SSS vertical variability determined from ARGO measurements

S5-10-S0-5 is small: <0.01 in 80% of cases but systematic differences with fresher S in

upper layer observed in the tropics

Boutin and Martin, 2006

Difference between 2 successive measurements on the same profile(S5-10 – S0-5) - July 2004-July 2005

0m

5m

10m

S0-5

S5-10

-0.02 0.10S5-10-S0-5

SSS bias as a function of wind speed bias (constant bias put on zonal wind speed)

Simulations of SSS retrieved from biased wind speeds

SSS bias mostly related to wind speed bias (at 15°C, 1psu bias <-> 2m/s bias); SSS bias generated by wind speed bias is slightly more important at low SST

WIND SPEED BIAS (m/s)

In order to get SSS bias<0.1psu , need for bias on wind speed data < 0.2m/s

SSS

BIA

S (

m/s

)

-8

8

-10 10WIND SPEED BIAS (m/s)

SSS

BIA

S (

m/s

)-8

8

-10 10

0°C<SST<10°C10°C<SST<20°C20°C<SST<30°C

SSS bias as a function of SST (SST bias=5°C)

Influence of SST bias on retrieved SSS

SSS bias strongly dependent on SST: almost no bias around 15°C; >0 biases at low SST and <0 biases at high SST

In order to get SSS bias < 0.1psu, need unbiased SST especially at low and high SST: -at SST=30°C: SST bias<0.5°C

-at SST=0°C: SST bias<0.3°C (extreme value!)

Est

ima

ted

– r

efer

enc

e S

SS

(ps

u)

-8

-6

-4

-2

02

4

6

8

10

2 7 17 12 22 27SST (°C)

Principle of SSS retrieval from L-band radiometer measurements

At L-band (microwave frequency: 1.4GHz; skin depth: 1cm) radiometric signal is mainly affected by ocean surface emission (atmosphere is almost transparent):

Tb = e . SST = (1-R) SSTTb : brightness temperature e : sea surface emissivity =R ( θ, SSS, SST, U…) =reflexion coefficient (may be deduced from Fresnel equations for a flat sea) 

R depends on sea water permittivity (SSS, SST), on sea surface roughness usually parametrized with U…

Main lobe solid angle

Antenna pattern

Sky

Atmospheric emission & absorption

Ocean brightness temperature

Apparent temperature distribution

2D SMOS Field of View: SSS retrieved from multiangular Tb

SSS in 1 pixel (~40x40km) retrieved from multiangular Tb measurements (~40)

Retrieval performed using iterative inverse method

Nb indt. Meas. in 40x40km pixel

Boutin et al., JAOT, 2004

Vertical variability between 0-5m and 5m-10m

Difference is small <0.01 in 80% of cases but systematic differences with fresher S in upper layer observed in the tropics

Effect of diurnal SST change in the surface ocean on Tb

In case of very low wind speed, solar heating may create a diurnal cycle of temperature at the ocean surface (warm layer effect; few meters depth).

Model of Fairall et al. (1996):if solar heating exceeds cooling by net longwave radiation and turbulent heat transfer => development of a warm layer

Forcing: ECMWF heat fluxes and wind speed

Stage T. Pastier, 2000

Case of moderate wind speed

1 day simulation (1 July 1999)

0 7 m/s

1°C0

ECMWF wind speed

T6PM –T6AM at 1cm depth

Tb_at_6PM –Tb_at_6AM

-0.1 0.1K

10 days simulation (1-10 July 1999)

Diurnal SST variation at 1cm depth: Significant at 6PM LT (descending orbits)

Over 10 days, T° at 6PM warmer by up to 1.5°C=> Tb variation between -0.2K (tropics-mid lat. in summer) and 0.1K (high latitude)

Weak but not negligible effect for SSS retrieval:

In the summer hemisphere,bias > 0.1psu in 14% of the rasters if diurnal SST effect is not corrected

Need to correct this effect when using evening orbit

measurements

T6PM –T6AM at 1cm depth

Tb_at_6PM –Tb_at_6AM0 1°C

-0.1 0 0.1K

Forward Modeling of SMOS apparent temperature

Apparent temperature measured by the SMOS radiometer affected by:

– the emission of the sea surface : • Klein and Swift model + tabulations of Dinnat et al. 2-scale model

– the emission and attenuation by the atmosphere • Liebe et al (1993) model

– the emission of the sky reflected at the sea surface• Reich maps + specular reflexion

– Ionospheric effects (Faraday rotation)• Waldteufel et al. 2004 parametrization

– Sunglint • Scattering deduced from SSA model (see N. Reul presentation)

Expansion of Tbsea in Fourrier serie (no swell): Tb=Tbn (SSS, SST,θ) +Tb0(U,θ) +Tb1(U,θ) cos(φ) +Tb2(U,θ) cos(2φ) Tbn : variations induced by SSS and SST for a flat surface (no wind) : Klein and Swift model

Forward Modeling of SMOS apparent temperature: Brightness temperature of the sea surface (1)

SST=0°C SST=15°C SST=30°C dTb/dSSS -0.22K/psu -0.45K/psu -0.65K/psu

SST=0°C SST=15°C SST=30°C dTb/dSST +0.1K/°C 0.005K/°C -0.16K/°C

Expansion of Tbsea in Fourrier serie (no swell): Tb=Tbn (SSS, SST,θ) +Tb0(U,θ) +Tb1(U,θ) cos(φ) +Tb2(U,θ) cos(2φ)

Tb0 : isotropic variations induced by the wind speed (Dinnat et al. model ; 2x Durden and Vesecki wave spectrum)

 Tv0 and Th0 as functions of wind speed at nadir, 30° and 50° incidence angles.

At nadir dTb0/dU 0.2K/m/s

Forward Modeling of SMOS apparent temperature: Brightness temperature of the sea surface (Dinnat et al. model) (2)

2 2.5 3 3.5 4 4.5 541

41.5

42

42.5

43

43.5

44

SPAIN

FRANCE

Over the Mediterranean Sea on November, 23rd, 2001 (no rain)

Validity of emissivity model ?? Comparison with Eurostarrs measurements

QSCAT WIND SPEED

Contardo et al, 2002Dinnat et al., 2002Coll. J. Miller, D. Long

Good agreement of Tb = Tbmeasured - Tb (SSS,SST)with Dinnat et al. roughness parametrization; scatter partly due to radiometer imperfections

Dinnat et al., 2002

0.27 K/m/s ; б = 1.12°K0.23 K/m/s ; б = 1.01°K

Signal induced by roughness at 2 incidence angles (21°) & (38.5°)

Validity of emissivity model ?? Comparison with Eurostarrs measurements

Thèse Dinnat, 2003Etcheto et al., TGARS, 2004