QWG8, Boutin et al.

SMOS and Aquarius: SSS and Wind Effect

J. Boutin, X. Yin, N. Martin (LOCEAN, Paris),

E. Dinnat (Chapman University/NASA/GSFC),

S. Yueh (NASA/JPL)

Coll. G. Reverdin (LOCEAN), G. Alory (LEGOS), F. Gaillard (LPO)

ESA Expert Support Lab.(ARGANS, ICM, Ifremer, ACRI-st, CLS)

ESA/CNES SMOS CAL/VAL project (GLOSCAL)

NASA ROSES project (Intercalibration of SMOS and Aquarius)

QWG8, Boutin et al.

Overview of this presentation

• SMOS – Aquarius – ARGO maps

• SMOS & Aquarius wind dependencies

• Comparisons of SMOS SSS wrt in situ measurements (ship and ARGO)

• Influence of using SSMI WS intead of ECMWF

QWG8, Boutin et al.

Data

Improvements with respect to previous versions:

-Reduced coastal bias; OTT (systematic bias in the Field of View) every 2 weeks and

separetely for ascending and descending orbits; improved wind correction ; improved

RFI & outlier sorting

SSS averages weighted by the variance of the error of the retrieved SSS

(Boutin et al., TGRS, 2012), and restricted to wind speed between 3 and

12m/s

USE ONLY ASCENDING ORBITS

SATELLITE SSS-SMOS SSS reprocessed by ESA v5.5 (wind-model 1)

-AQUARIUS level 3 maps v1.2.3

IN SITU SSS

-Optimal Interpolation of ARGO SSS: IFREMER In Situ Analysis System (ISAS)

(Gaillard et al. 2009)

-ARGO SSS (upper SSS between 10m and 0.5m depth) (CORIOLIS GDAAC)

-Ships of Opportunity (Delayed mode data delivered by SSS observatory

(http://www.legos.obs-mip.fr/observations/sss/))

QWG8, Boutin et al.

Ascending vs Descending SSS: Sept 2011

SSS Ascending orbits SSS Descending orbits

SSS error Ascending orbits SSS error Descending orbitsDescending OTTs not enough to remove stripesespecially in September=> Sun alias? Galactic noise?

QWG8, Boutin et al.

Sept-Dec 2011

SMOS v5.5(Ascending Orbits)

Aquarius v1.2.3

ARGO OI

SMOS - AQUARIUS – ARGO SSS – Sept-Dec 2011

QWG8, Boutin et al.

SMOS-Aquarius

SMOS-ARGO OI

Aquarius-ARGO OI

Sept-Dec 2011

SMOS features: -issues close to land (image reconstruction) -Large RFI (N. Atl., Asian coasts) -Ice edge

Aquarius features: -Underestimates in Southern Ocean -Overestimates close to islands (e.g. S. Pac ).

SMOS & Aquarius features: -Better resolution of Amazone Plume -Lower SSS in rainy regions -Lower SSS in Indian Ocean

-1 +1

-1 +1

-1 +1

QWG8, Boutin et al.

SMOS & AQUARIUS WIND DEPENDENCIES

Aquarius Tb-wind :

-empirical wrt SSMI wind speed + NCEP wind direction (Yueh)

-empirical wrt NCEP wind (Dinnat)

SMOS Tb-wind :

-Adjusted rough & foam models wrt SSMI wind speed (Yin et al, TGRS, 2012)

-Adjusted rough & foam models wrt ECMWF wind speed (Yin et al, TGRS, 2012)

-Empirical model wrt ECMWF wind speed (Tenerelli, 2011)

-Empirical model wrt ECMWF wind speed (Guimbard et al, 2012)

Given the uncertainty about the absolute bias between model and data (need for OTT

correction for SMOS) all the curves have been adjusted at 7m/s

QWG8, Boutin et al.

Very similar SMOS and Aquarius Tb-Wind omnidirectional dependencies 3<WS<12 m/sAbove 15m/s SMOS T higher than Aquarius T (especially in H polarization)

Aquarius (SSMI WS); Aquarius (NCEP WS); SMOS (SSMI WS); *,*,* SMOS (ECMWF WS)

Omnidirectional Tb-Wind dependencies

28.7° 37.8° 45.6°

Th(K)

Tv(K)

7

7 7 7

77

00

0 0 0

0

2 17WS (m/s) WS (m/s) WS (m/s)2 17 2 17

2 172 172 17WS (m/s) WS (m/s) WS (m/s)

QWG8, Boutin et al.

Tb dependency with Wind Direction

0 20WS (m/s) WS (m/s) WS (m/s)0 20 0 20

28.7° 37.8° 45.6°0.9

0.6

0.3

0 -0.3

T(K)

0.9

0.6

0.3

0 -0.3

0.9

0.6

0.3

0 -0.3

Aquarius

SMOS model 1

V1V2H1H2

Tb=Tbwind + Tb1 cos(Phi) + Tb2 cos(2Phi)

TbV1 TbV2 TbH1 TbH2

Much larger dependency in Aquarius than in SMOS functions at high WS!

SMOS and Aquarius Tb wind dependencies better agree for wind speed between 3 and 12m/s

=> confirm the relevance of SMOS L2 wind speed range flagging

QWG8, Boutin et al.

Comparisons of SMOS reprocessed SSS(v5) and in situ SSS

• ARGO

• Ship:– Indian Ocean (Lavender)– Atlantic Ocean:

• Rio Blanco• Colibri and Toucan

QWG8, Boutin et al.

Comparisons with ARGO in selected regionsfar from land (Aug-Sept 2010 & 2011)

Atl subtrop

Ind. S

Pac S

Pac trop N

32

40

4032

(SMOS SSS weighted averaged over 100km-10days around ARGO)SMOS – ARGO SSS in tropical Pacific 0.1 fresher than in other regions: rain?

SS

Ss

mo

s

SSSargo

QWG8, Boutin et al.

Indian Ocean: SMOS–Lavender Ship SSS19 Sep – 1 Oct 2010

Sept 2010 - SMOS SSS (Asc orbits) SMOS – ARGO OI SSS

Next slides: SMOS SSS averaged over 10 days and 100km around Lavender time and location

QWG8, Boutin et al.

ShipSMOS 10d-100kmARGO OI

ARGO OI much smoother than Ship SSS: monthly & ~1 ARGO meas./10days, 2°

SMOS (40km resolution) better sees small scale variability, but still some deficiencies (distance to coast? , RFI, islands…)

On average:SSSARGO_OI-SSSTSG = 0.2 (0.23)SSSSMOS-SSSTSG = 0.0 (0.37)

SSS latitudinal profiles – Ship-ARGO OI-SMOS

QWG8, Boutin et al.

Atlantic Ocean SSS: SMOS and Rio Blanco Ship (Aug-Sep 2010)

4030-40

50

Distance to coastSSS

ITCZ

High noise close to land but qualitative agreementSMOS SSS fresher than ship in ITCZ

Latit

ude

ShipSMOS 10d-100km

QWG8, Boutin et al.

Comparison of Toucan and Colibri ship and SMOS v5 reprocessed SSS

Leslie David (master trainee), Gilles Reverdin, Jacqueline Boutin

(LOCEAN)

QWG8, Boutin et al.

SPURS region

Toucan and Colibri Ship of Opportunity trajectories (2010-2011)

QWG8, Boutin et al.

32

38

ARGO interpolation

SMOS Ascending 10days-100kmSMOS Ascending 10days-100km

Colibri Ship

SSS

Latitude (°)5N 35N

18-22 October 2011

QWG8, Boutin et al.

QWG8, Boutin et al.

Mean and std of difference (SMOS-Ship SSS) over the whole period 2010-2011and July-Aug-Sept

Coast/RFIBiases?

Std ~0.2-0.3 (except north of 25N in winter-Spring; RFIs?)

(all sections except 19-Oct-2010, 11-Jul-2010 et 14-Mar-2011 )

10N 15N 20N Latitude 30N 35N

-0.2

0.2

1

-1

Bia

s an

d st

d(di

ff)

Black /grey: all monthsColors: July-Aug-Sept

QWG8, Boutin et al.

RFI sorting needs to be improved!!

35.5

38.5

16N 34N

June 2011

QWG8, Boutin et al.

Conclusion

SMOS SSS:• In tropical-subtropical regions, far from land, in non rainy region, precision

at 100x100km2-10days ~0.3

• SMOS SSS freshening wrt in situ SSS (~5m depth) in rainy conditions: true salinity effect? Needs for surface drifters for validation (needs for a lot of matchups!!!)

• Ship TSG: very useful for validating ‘small scale’ (<200km) variability (gradients) seen by SMOS

• Still issues (sun aliases, land vicinity, ice edge, RFI sorting …): image reconstruction still in progress

SMOS and Aquarius wind models

• Tb-wind SMOS and Aquarius very similar: 3-12m/s but Aquarius lower above 12m/s; Aquarius azimuth dependency much larger at high wind speed

QWG8, Yin et al.

Issues about wind speed uncertainties and SMOS retrieved SSS

Yin X.1, Boutin J.1, Martin N.1, Vergely J.2, Spurgeon P.3

1. LOCEAN/IPSL, UMR CNRS/UPMC/IRD/MNHN, France 2. ACRI-ST, France 3. ARGANS, United Kingdom

QWG8, Yin et al.

OutlineImpact of wind speed uncertainties on SMOS retrieved SSS

1) SSMI versus ECMWF wind speed

2) Difference in retrieved SSS using two different prior wind speeds

3) The eastern equatorial Pacific Ocean

Conclusions and perspectives

QWG8, Yin et al.

Wind speed retrieved from SMOS data

August 2010, ECWMF wind speed used as prior.

-300 300

WS (retrieved-prior)

Distance to the track (km)

Latitude

40N

50S

Incidence angle

60

0

Eta

Distance to the track (km)

In the centered swath (within +-300km), wind speed is adjusted with SMOS data; in the border of the swath (outside of +-300km), wind speed can not be adjusted with SMOS data (not enough incidence angles).

QWG8, Yin et al.

Data

1. SMOS Brightness temperature (last reprocessing: L1c v504)1) August 2010, 436 ascending orbits over global ocean2) September 2011, 418 ascending orbits over global ocean

2. ECMWF wind speed collocated with SSMI wind speed,August 2010 and September 2011collocation radii of ±0.5 hour and ±50 km

3. SSS1) SMOS SSS retrieved with ECWMF wind as prior. Roughness model fitted from ECMWF wind speed (Yin et al., TGRS SMOS special issue, 2012) is used. ( ESA reprocessing v550)

2) SMOS SSS retrieved with SSMI wind as prior, processed by LOCEAN. Roughness model fitted from SSMI wind speed (Yin et al., TGRS SMOS special issue, 2012) is used.

3) ARGO optimal interpolation (ISAS) (Gaillard et al. 2009) used as reference

QWG8, Yin et al.

For the orbit 20100809T142544_20100809T151945, the ECMWF and SSMI wind speeds are quite different around 17S.

One example of differences between ECMWF, SSMI and retrieved WS from SMOS data

QWG8, Yin et al.

Latitude

TB simulations with SSMI Wind

5 points running mean of SMOS TB

TB simulations with ECMWF Wind

No matter whether SSMI or ECWMF wind is used as prior, retrieved wind speed is closer to SSMI wind speed but still a little bit lower.

TB simulations with SSMI Wind are closer to SMOS TB

_____ECMWF prior WS---------SSMI prior WS• • Retrieved WS (ECMWF prior)• • Retrieved WS (SSMI prior)

QWG8, Yin et al.

August 2010 monthly averageSMOS SSS (+-300km)3-12m/s (both ECWMF and SSMI)

Similar spatial distribution

The same for September 2011, not shown here.

SMOS SSS with SSMI wind speed

SMOS SSS with ECMWF wind speed

ARGO OI SSS

QWG8, Yin et al.

August 2010 monthly averageSMOS SSS – ISAS/ARGO SSS3-12m/s

1. In the eastern equatorial Pacific ocean, differences between SMOS SSS with SSMI WS and ARGO OI SSS are often lower than those with ECMWF WS.

2. In the eastern part of Pacific and Atlantic ocean, SSMI WS are lower than ECMWF WS. Also, the retrieved SSS with SSMI WS are lower than those with ECMWF WS.

Similar patterns in September 2011, not shown here.

SMOS– ARGO OI SSS, SSMI prior wind speed

SMOS– ARGO OI SSS, ECMWF prior wind speed

1

-1

1

-11

-11

-1

SMOS SSS (SSMI) – SMOS SSS (ECMWF)

WS (SSMI) – WS (ECMWF)

QWG8, Yin et al.

Retrieved WS(ECMWF)

Retrieved WS(SSMI)

rWS(ECMWF)- ECMWF WSrWS(SSMI)- SSMI WS

-Monthly averaged August 2010-

1. In the eastern equatorial Pacific ocean, SSMI WS are much lower than ECMWF WS. 2. Retrieved WS with ECMWF WS are trying to get close to SSMI WS but not close enough.

QWG8, Yin et al.

Conclusions and Perspectives

1. In the centered part of the swath (within +-300km), wind speed is adjusted with SMOS data; In the border of the swath (outside of +-300km), wind speed can not be adjusted with SMOS data.

2. In the centered part of the swath (+-300km), the difference between SMOS SSS and ARGO OI SSS is lower than the border, since wind speed is corrected (partly) with SMOS data.

3. Locally (eastern equatorial Pacific), SSMI wind speeds used as prior in the retrieval perform better (SMOS retrieval scheme not able to completely correct for the differences between SSMI and ECMWF WS, but it decreases SSS bias locally).

4. Origin of differences in eastern equatorial Pacific?

5. Work in progress….

QWG8, Yin et al.

Sensitivity of L-band Tb (h- & v-pol) to θ, wind speed and SSS

Incidence angle

Incidence angle

From Dinnat et al. 2003

Incidence angle (°)

Tv/ WS >0 and with incidence angle

Tv/ SSS<0 and abs() with incidence angle

Th/ WS>0 and with incidence angle

Th/ SSS<0 and abs() with incidence angle

(at nadir, 20°C, Tb/ WS ~0.2K/m/s, Tb/ SSS~-0.5K/psu => a bias of 5m/s on WS ~ a bias of -2psu on SSS)

SMOS SSS (ESA level 2 retrieval) is retrieved through a least square minimisation of the difference between SMOS and modeled Tb.

1

02

2,

1

02

2mod2 ),( Np

j P

priorjjNm

i T

bimeas

ib

jib

PPPTT

Retrieval of SSS (=100psu), SST (=1°C), WS(=1.5m/s on wind components) through the minimisation of:

=> estimate of SSS error (Levenberg & Marquard algorithm)

SMOS SSS and WS retrieval method

QWG8, Yin et al.

SSS v5 Descending August & September 2010 & 2011

August 2010

September 2011August 2011

September2010

Descending OTTs not enough to remove stripesLarger stripes in September than in August=> Sun alias? Galactic noise?

Less outliers in the northern Atlantic in 2011 than in 2010Decrease of RFIs