Sea Surface Salinity Retrieval: Preliminary results J ...earth.esa.int/smos04/posters/boutin.pdf · SMS study 5th SMOS workshop Sea Surface Salinity Retrieval: Preliminary results

SMS study 5th SMOS workshop

Sea Surface Salinity Retrieval: Preliminary results

J. Boutin, J.L. Vergely, P. Waldteufel

Objective:Assess sensitivity of SSS retrieval to operation mode

(dual pol versus full pol), to errors on auxiliary parameters (SST, wind, TEC)


Tests conducted over half an orbit; central Pacific ocean

Ascending orbit Descending orbit (TEC tests)


Geophysical data and direct model used to simulate Tb

SSS: -Levitus climatology (January month)SST: -Reynolds SST (30 Dec 2001-5 Jan. 2002)Wind components: -QSCAT daily wind speed (02 January 2002)TEC (N. Floury pers. comm.): -IRI model (Jan. 1989 (high solar activity))

Ocean emission: Tbocean=Tbflat + Tbrough� Tbflat: -Klein & Swift model � Tbrough: -Tabulations derived with the Dinnat et al. 2-

scale emissivity model (2xDurden & Vesecki wave spectrum; no foam)


Data used for Tb simulations: TEC (TecU)

6AM 6PM


SSS retrieval conditions (1)

-Blackman window-Tscene = 110K (preliminary value; definitive value taken as 90K);-Tsystem=180K (preliminary value; definitive value taken as 215K)-grid: 40km equidistant nodes => independent measurements-Iterative inversion scheme-a-priori SSS: 35; infinite error on a-priori SSS-Auxiliary parameters : wind components, SST,TEC-Input values deduced from true values + gaussian noise


SSS retrieval conditions (2)

σSST: 0;0.5;1;1.5,2;2.5;3σVx= σVy=0,1,1.5,2,3,4,5

σSST: 1°CσVx=σVy=1.5m/sσTEC=5TecU

Ascending, Pacific

DP,FP, ST1D, ST1F

A priori error (sensitivitytests)

A priori error (default)orbitPolarizationmode

DP: SSS retrieved from A1,A2; radiometer in dual pol modeFP: SSS retrieved from A1,A2,A3,A4; radiometer in full pol mode; σA1, σA2=√3/2 σA1, σA2 in DPST1D: SSS retrieved from (A1+A2); radiometer in dual pol modeST1F: SSS retrieved from (A1+A2); radiometer in full pol mode

-Inversion of antenna temperatures, Ai:A1=c2 T1 +s2 T2A2= s2T1 +c2 T2A3=s2T1 -s2 T2A4=0

Where T are surface brightness temperatures, T3 and T4 are assumed to be 0, c=cos(α) and s=sin(α) and (α) is an angle depending on the satellite geometry and on the Faraday rotation angle (Waldteufel et al., 2004)


SSS errors: Dual Pol versus Full Pol; 2pol versus Stokes 1 retrievals (default conditions)

Use of A1,A2:Full pol more homogeneous over theswath than dual pol (better correction for Faraday rotation sensitive at highincidence angles); DP better than FP at the center of theswath (low incidence angles; lowererror on Tx,Ty)

Use of St1:DP better than FP everywhere (lowererror on Tx,Ty)

SSS error averaged over all latitudes


Sensitivity tests: influence of SST errors on SSS errors

σSST=1°C σSST=2°C

Errors more homogeneous in FP than in DPFP more robust than DP to an increased error on SST

DP; FP; ST1DP; ST1FCumulated probability density of SSS errors for the 4 modes



Cumulated probability density of SSS errors for the 4 modes

σSST=1°C σSST=3°C

Errors more homogeneous in FP than in DPFP more robust than DP to an increased error on SST

DP; FP; ST1DP; ST1F



Cumulated probability density of SSS errors for variable SST errors: 0; 0.5; 1; 1.5; 2 °C etc

DP FP

SSS errors very similar for σSST=0. °C and σSST=0.5°C; No need for SST precision better than 0.5°C



Cumulated probability density of SSS errors for variable SST errors: 0; 0.5; 1; 1.5; 2 °C etc

ST1DP ST1FP

Same conclusion as before: SSS errors very similar for σSST=0.°C and σSST=0.5°C; No strong constraint for SST precision better than 0.5°C


SSS errors averaged over all latitudes: DP;FP;ST1DP;ST1FP

For low wind errors (<1m/s):inversion of (A1,A2) or St1 are equivalent; DP mode better than FP mode

FP mode is less sensitive to increasing wind errors than DP mode

σUx= σUy=0m/s σUx= σUy=1m/s

Sensitivity tests: influence of wind errors on SSS errors


SSS errors averaged over all latitudes : DP;FP;ST1DP;ST1FP

For wind components errors above 1.5m/s:In DP, errors on SSS retrieved from (A1,A2) slightly lower than from St1FP mode better than DP mode

σUx= σUy=1.5m/s σUx= σUy=2m/s



DP FP

Cumulated probability density of SSS errors for variable wind component errors : σVx=σVy=0;1;1.5;2

FP less sensitive to wind components error than DP =>SSS errors lower in DP than in FP for V errors<1.5m/sSSS errors higher in DP than in FP for V errors>1.5m/s



ST1DP ST1FP

Cumulated probability density of SSS errors for variable wind component errors : σVx=σVy=0;1;1.5;2


Errors in ST1FP higher than in ST1DP because of higher radiometric noise


idem

idem

σSST: 1°CσVx=σVy=1.5m/s

A priori error on SST and Vx,Vy

TEC bias =[0, 2, 5, 10, 15, 20, 25, 30] (tecu) with TEC inversion

Ascending orbitDP

TEC bias =[0, 2, 5, 10, 15, 20, 25, 30] (tecu) without TEC inversion

Ascending orbitDP

σTEC=[0, 2, 5, 10] error (tecu)ascending and descending orbits

DP, FP, ST1D, ST1F

Inversion configurationorbitPolarizationmode

Sensitivity tests: influence of TEC errors and biases on SSS errors


Cumulative probability of SSS errors for variable TEC errors: 0;2;5;10;20(with TEC inversion)

DP: inversion does not retrieve TEC but low impact on SSS errorFP: above 5TecU error, similar SSS error distribution

Sensitivity tests: influence of TEC errors on SSS errors

DP FP


Cumulative probability of SSS errors for variable TEC errors: 0;2;5;10;20

ST1 not sensitive to TEC!!!


ST1DP ST1FP


SSS errors averaged over all latitudes : DP;FP;ST1DP;ST1FP(with TEC inversion)

TEC retrieval better in FP than in DP (of course!) : critical area: center of the swath; TEC error does not strongly affect SSS errors provided TEC is inverted (SSS errorssimilar to the ones obtained for σTEC=5TecU)


σTEC=20TecU

SSS error TEC error


SSS bias averaged over all latitudes : DP;FP;ST1DP;ST1FP

TEC bias does not strongly affect SSS errors provided TEC is inverted:It is essential to invert TEC

Sensitivity tests: influence of TEC biases on SSS

σTEC=20TecU

SSS bias; TEC not inverted SSS bias: TEC inverted


Preliminary conclusions

Preliminary simulations indicate:

FP mode => errors are more homogeneous across the swath⇒ More robust to SST and wind errors than DP:

FP mode better than DP for σU>1.5m/s and σSST>1°C

Inversion of (A1,A2) equivalent to inversion of ST1 if wind perfectly known, slightly better for σU>1.5m/s

TEC error and bias are not a big deal provided that TEC is inverted

Results very dependent on radiometric noise

Future simulations needed to assess effects of biases on SST, U


Ocean (constant values in order to ease results interpretation):SSS: 34.5psuSST: 15°CWind components: 7m/s

Land:Soil moisture= 0.15 m3.m-3 (error=0.1m3.m-3)Optical thickness assumed to be perfectly known except herbaceous vegetetion

known with an error of 30%Land temperature = 15°CLand cross polarization coefficient= 1 (a priori error =1)

Influence of coast proximity

Gulf of Bicay30 days (24km pixels)


Influence of coast proximity

Tested configurations

Apodization windows

Polarization mode

map orbit configuration

rectangular FP and ST1D

Coast data 30 days (ascending and descending)

per case

Retrieval of the sea and land parameters : SSS, SST, u, v, WS, τ,

Cpol

Blackman idem idem idem idem

Kaiser 13 idem idem idem idem


SSS error (24km � 1month);

kaiser 13 window, dual pol Pixels with 95% of water almost unaffected by the coast

Documents

Sea Surface Salinity Retrieval: Preliminary results J ...earth.esa.int/smos04/posters/boutin.pdf · SMS study 5th SMOS workshop Sea Surface Salinity Retrieval: Preliminary results