On SMOS SSS biases in 2012 ... SMOS – In situ comparisons in SPURS region
J. Boutin*, N. Martin*, O. Hernandez*, G. Reverdin*
*LOCEAN
Mean SSS in OTT region(2012: real time SSS; OTT not well phased)
N. Martin e al. 2012
2012
What differences between:- using a bias correction method based on observed SSS biases in
the ‘OTT region’ wrt to updating OTT
- using WOA climatology as reference or ISAS (interpolated in situ measurements) and using whole ‘OTT area’ (45°S-5°S) or a reduced area (30°S-5°S)
Mean SSS in OTT region(2012: real time SSS; OTT not well phased)
2012
Before the catch-up reprocessing, we test an adhoc correction (OCOR):
For each point in the OTT area (140W-95W; 45S-55S):DSSS=SSSsmos-SSSwoa => DTb(DSSS, SSS, SST, WS) => <DTb>ott
In each ocean pixel:DSSS=f(<DTb>ott, SSS, SST, WS)
Adhoc correction (OCOR):
For each point in the OTT area (140W-95W; 45S-55S):DSSS=SSSsmos-SSSwoa => DTb(DSSS, SSS, SST, WS) => <DTb>ott
In each ocean pixel:DSSS=f(<DTb>ott, SSS, SST, WS)
Daily <DTb>ott Ascending orbits
5 day running mean of <DTb>ott Ascending orbits
Various curves using WOA SSS or ISAS SSS, 5S-30S or 5S-45S as reference=> weak differences (<0.1K but not completely negligible)
Daily <DTb>ott very noisy => we use 5day running means computed on ascending and descending orbits separetly and using WOA as reference
WOA, 5S-45SISAS, 5S-45SWOA, 5S-30SISAS, 5S-30S
Adhoc correction (OCOR):
For each point in the OTT area (140W-95W; 45S-55S):DSSS=SSSsmos-SSSwoa => DTb(DSSS, SSS, SST, WS) => <DTb>ott
In each ocean pixel:DSSS=f(<DTb>ott, SSS, SST, WS)
Daily <DTb>ott Descending orbits
5 day running mean of <DTb>ott Descending orbits
Various curves using WOA SSS or ISAS SSS, 5S-30S or 5S-45S as reference
Daily <DTb>ott very noisy => we use 5day running means computed on ascending and descending orbits separetly and using WOA as refernce
WOA, 5S-45SISAS, 5S-45SWOA, 5S-30SISAS, 5S-30S
BIASES on monthly SSS maps with respect to ISAS SSSAscending orbits Descending orbits
OTT Area
45°N-45°S
Less variable biases with OCOR than with REPR because of daily correction
‘Global’ biases very different from biases in OTT region!!
REPR-ISASWOA-ISAS
OCOR-ISAS
Different biases in OTT region and other regions......
STD of (SMOSOCOR SSS - SMOSREPR SSS)Ascending orbits Descending orbits
45°N-45°S
The temporal variation of biases within 15days + the variation of the biases in the FOV => std >0.1 on monthly 1° maps!!!
Almost no difference between applying a mean daily bias (OCOR_1day) or a 5 days running mean (OCOR_5days)
STD of (SMOS SSS - ISAS SSS) for OCOR and OTT_Repr correctionsAscending orbits Descending orbits
OTT Area
45°N-45°S Although OTT_REPR correction is not daily, std (SMOS-
ISAS) is equivalent (or even smaller) for REPR than for OCOR => a mean correction is less efficient than a varying OTT correction (at least with uncorrect galactic map!)
OCOR-ISAS
REPR-ISAS
October 2012 Desc.
Over 45S-45N :OCOR-ISAS : -0.17 (std=0.47)
REPR-ISAS : 0.28 (std=0.45)
SummaryA correction deduced from mean SSS bias over ‘OTT area’ corrects a large part of biases but remains less efficient than updating OTT
Needs for updating OTT more frequently than every 2 weeks will improve
But
Large north-south biases remain
In ‘OTT region’ : North-south variations of biases and SSS variations wrt climatology responsible for uncertainty in OTT <0.1K
SPURS area
Large scale bias between SMOS & AQUARIUS averaged over SPURS area
We first correct for a mean bias between SMOS & WOA or AQUARIUS & WOA
Comparisons SMOS-Toucan & Colibri(+/-9jours, +/-50km)
Mean and standard deviation of SSS SMOS – SSS TSG for each campaign (Colibri and Toucan). Correction of bias B1 is applied. In green: colocalisations in October 2011
Mean negative bias because WOA SSS often lower than in situ SSS
SSS in October 2011 in SPURS regionSMOS SSS theoretical error
SSS in October 2011 in SPURS region
SSS in August 2012 during the STRASSE campaign
Boutin et al., AGU 2012
All data
R2 =0.78n = 30896 mean = -0.06 std = 0.2median = -0.04 RMSE = 0.21
Error < 0.4 PSS-78
R2 =0.81n = 23835 mean = -0.03 std = 0.18median = -0.02 RMSE = 0.19
Summary of SPURS comparisons
Once large scale seasonal bias has been corrected, rms error of SMOS SSS ~0.2; SMOS allows a good monitoring of SSS spatial structures at 50-100km scale