Post on 17-Jan-2016
SMOS Quality Working GroupMeeting #2
Frascati (Rome), September 13th-14th,2010
SMOS-BEC Team
2 / Ntot
Outline
• METHODOLOGY• TESTS
1. RETRIEVAL MODE
Dual from Full vs. Stokes from Full
2. BIAS MITIGATION
No correction vs. External Bias Temperature Calibration vs. OTT
3. MODELS
Model 2 vs. Model 3(16)
4. SSS SELECTION
All overpasses vs. Ascending vs. Descending
5. TB SELECTION
EAF vs. AF
6. NEW FTR
July vs. August• CONCLUSIONS (+ or -)
3 / Ntot
Methodology
All the results presented are at Level 3 (10-day 2-degree product).
Retrievals have been performed using SMOS-OS Level2 Processor.
Level 2 data have been filtered according to:
Fg_ctrl_reach_maxiter1,2,3: Maximum number of iteration reached before
convergence.
Fg_ctrl_marq1,2,3: Iterative loop ends because Marquardt
increment is greater than lambdaMax (100).
Statistical characterization is done considering only points more than 200 km from the coast
Fg_sc_land_sea_coast1 = 1 & Fg_sc_land_sea_coast2 = 0
4 / Ntot
Methodology
L2 L3 averaging has been performed according to:
iobsi
iobsiiL
L N
NSSSSSS
2
3
The L3 accuracy is also introduced to someway estimate the quality of the measurement
iobsi
iobsiiLiL
L N
NSSSSSSACC
2
22
3
5 / Ntot
Tests
1.RETRIEVAL MODE
10 days of retrieval from July, 10th to 19th
ISEA4H9 ISEA4H8 to reduce the computational resources needed
Model 2 in the mode “Stokes from Full-Pol” has been used
OTT has been applied in accordance to the official DPGS product
L3 retrieved SSS is compared to NOAA WOA05 climatology and ARGO averaged data
6 / Ntot
Tests – Dual vs. Stokes’ I
L3 maps - Dual
Amazon plume
Cold waters
7 / Ntot
Tests – Dual vs. Stokes’ I
L3 maps - Stokes’ I
Amazon plume
Cold waters
8 / Ntot
Tests – Dual vs. Stokes’ I
L3 maps
9 / Ntot
Tests – Dual vs. Stokes’ I
L3 maps - Accuracy
2.5 psu
10 / Ntot
Tests – Dual vs. Stokes’ I
L3 statistics - Dual
rms
0.2260 2.2565 2.2678
0.5559 1.4457 1.5489
rms
0.2958 1.4391 1.4692
0.5278 0.5805 0.7846
11 / Ntot
Tests – Dual vs. Stokes’ I
L3 statistics – Stokes’ I
rms
0.2445 2.3505 2.3631
0.6157 1.5464 +7%
rms
0.2935 1.6756 1.7011
0.6080 0.5932 0.8495
+4%
1.6645
+14%
+8%
12 / Ntot
Tests
2.BIAS MITIGATION
10 days of retrieval from July, 10th to 19th
ISEA4H9 ISEA4H8 to reduce the computational resources needed
Model 2 in the mode “Dual from Full-Pol” has been used
No correction, external brightness temperature calibration[*], and OTT have been applied
L3 retrieved SSS is compared to NOAA WOA05 climatology and ARGO averaged data
13 / Ntot
Tests – Bias mitigation
External Brightness Temperature Calibration
Constant within the snapshot (xi, eta) but varying in time
Ocean Target Transformation
Constant in time but varying within the same snapshot
snapshotmodretcorr TBTBTBTB
),( modretcorr TBTBTBTB
14 / Ntot
Tests – Bias mitigation
L3 maps – No bias mitigation
15 / Ntot
Tests – Bias mitigation
L3 maps – External Brightness Temperature Calibration
16 / Ntot
Tests – Bias mitigation
L3 maps – External Brightness Temperature Calibration
MEAN BIAS SUBTRACTED
Less intense land-sea transition effect
17 / Ntot
Tests – Bias mitigation
L3 maps – Ocean Target Transformation
18 / Ntot
Tests – Bias mitigation
L3 maps - Accuracy
2.5 psu
19 / Ntot
Tests – Bias mitigation
L3 statistics – no bias mitigation
rms
3.9591 2.6742 4.7776
3.3468 3.7604 5.8288
rms
4.1488 1.8571 4.5455
3.5769 0.6343 3.6327
20 / Ntot
Tests – Bias mitigation
L3 statistics – External Brightness Temperature Calibration
rms
2.9600 2.3886 3.8036
2.1981 1.7300 2.7972
rms
3.1541 1.7764 3.6199
2.2214 0.5801 2.2959
-14% -20%
-52% -37%
21 / Ntot
Tests – Bias mitigation
L3 statistics – Ocean Target Transformation
rms
0.2260 2.2565 2.2678
0.5559 1.4457 1.5489
rms
0.2958 1.4391 1.4692
0.5278 0.5805 0.7846
-52%
-73%
-68%
-78%
22 / Ntot
Tests
3.MODELS
10 days of retrieval from July, 10th to 19th
ISEA4H9 ISEA4H8 to reduce the computational resources needed
OTT has been applied as for the official DPGS product
Model 2 and Model 3(16) are compared
L3 retrieved SSS is compared to NOAA WOA05 climatology and ARGO averaged data
23 / Ntot
Tests – Model 2 vs. Model 3(16)
L3 maps – Model 2
24 / Ntot
Tests – Model 2 vs. Model 3(16)
L3 maps – Model 3(16)
25 / Ntot
Tests – Model 2 vs. Model 3(16)
L3 maps
26 / Ntot
Tests – Model 2 vs. Model 3(16)
SST L3 maps
27 / Ntot
Tests – Model 2 vs. Model 3(16)
WS L3 maps FROM ASCAT
28 / Ntot
Tests – Model 2 vs. Model 3(16)
Scatterplot
29 / Ntot
Tests – Model 2 vs. Model 3(16)
L3 maps - Accuracy
2.5 psu
30 / Ntot
Tests – Model 2 vs. Model 3(16)
L3 statistics – Model 2
rms
0.2445 2.3505 2.3631
0.6157 1.5464 1.6645
rms
0.2935 1.6756 1.7011
0.6080 0.5932 0.8495
31 / Ntot
Tests – Model 2 vs. Model 3(16)
L3 statistics – Model 3(16)
rms
0.8302 2.5735 2.7041
0.8490 1.7188 1.9171
rms
0.9346 1.9486 2.1611
0.9522 0.6619 1.1597
+13%
+13%
+21%
+27%
32 / Ntot
Tests
4.SSS SELECTION
10 days of retrieval from July, 10th to 19th
ISEA4H9 ISEA4H8 to reduce the computational resources needed
OTT has been applied as for the official DPGS product
L3 averaging has been performed using ALL the overpasses, only the ASCENDING ones, and only
the DESCENDING ones
L3 retrieved SSS is compared to NOAA WOA05 climatology and ARGO averaged data
33 / Ntot
Tests – All vs. Ascending vs. Descending
L3 maps - All
34 / Ntot
Tests – All vs. Ascending vs. Descending
L3 maps - Ascending
Fresher when ice/land enters in the FOV
Saltier when it exits
35 / Ntot
Tests – All vs. Ascending vs. Descending
L3 maps - Descending Generally saltier
Fresher when ice/land enters in the FOV
Saltier when it exits
36 / Ntot
Tests – All vs. Ascending vs. Descending
L3 maps – comparisons with Ext TB cal
Tests – All vs. Ascending vs. Descending
land-sea contamination
a previous study
38 / Ntot
Tests – All vs. Ascending vs. Descending
L3 statistics - All
rms
0.2260 2.2565 2.2678
0.5559 1.4457 1.5489
rms
0.2958 1.4391 1.4692
0.5278 0.5805 0.7846
ALL PASSES ALL PASSES
39 / Ntot
Tests – All vs. Ascending vs. Descending
L3 statistics - Ascending
rms
-0.0214 3.3183 3.3183
0.2175 1.5405 1.5557
rms
-0.3585 3.0150 3.0362
0.1239 0.8506 0.8595
+32%
=
+52%
+9%
40 / Ntot
Tests – All vs. Ascending vs. Descending
L3 statistics - Descending
rms
0.3851 2.9766 3.0014
0.9942 1.7178 1.9848
rms
0.7824 1.8510 2.0096
1.0820 0.9610 1.4472
+24%
+22%
+27%
+46%
41 / Ntot
Tests
5.TB SELECTION
5 days of retrieval from July, 10th to 14th
ISEA4H9 has been used
Model 2 in the mode “Dual from Full-Pol” is analized
OTT has been applied as for the official DPGS product
TB with a have been filtered out to almost reproduce the AF-FOV
L3 retrieved SSS is compared to NOAA WOA05 climatology and ARGO averaged data
25.022
42 / Ntot
Tests – EAF vs. AF
AF-FOV approx.
43 / Ntot
Tests – EAF vs. AF
L3 maps - EAF
44 / Ntot
Tests – EAF vs. AF
L3 maps - AF
45 / Ntot
Tests – EAF vs. AF
L3 maps – AF minus EAF
Ascending positive
Descending negative
46 / Ntot
Tests – EAF vs. AF
L3 statistics - EAF
rms
-0.1680 2.7191 2.7242
0.4170 1.1453 1.2189
rms
0.0068 1.8762 1.8762
0.3608 0.6828 0.7722
47 / Ntot
Tests – EAF vs. AF
L3 statistics – AF
rms
-0.0867 2.8017 2.8030
0.5254 1.2677 1.3722
rms
0.0674 1.9639 1.9650
0.4797 0.7427 0.8841
+3%
+11%
+5%
+13%
48 / Ntot
Tests
6.NEW FTR
10 days of retrieval from July, 10th to 19th and August, 20th to 29th are compared
as produced by the DPGS:
ISEA4H9 has been used
Model 2 in the mode “Dual from Full-Pol” is analyzed
OTT has been applied
L3 retrieved SSS is compared to NOAA WOA05 climatology and ARGO averaged data
49 / Ntot
Tests – July vs. August
L3 maps - July
50 / Ntot
Tests – July vs. August
L3 maps - August Generally fresher
51 / Ntot
Tests – July vs. August
August minus July
52 / Ntot
Tests – July vs. August
L3 statistics - July
rms
-0.0308 2.6585 2.6587
0.5923 1.0897 1.2403
rms
0.1905 1.7765 1.7867
0.5401 0.5574 0.7762
53 / Ntot
Tests – July vs. August
L3 statistics – August
rms
-0.4285 2.8671 2.8990
0.2659 1.1050 1.1366
rms
-0.2197 1.8899 1.9027
0.2691 0.5818 0.6410
+8%
-8%
+6%
-17%
Conclusions (+ or -)
1. Dual from Full vs. Stokes from Full
4-14 % increment in SSS misfit rms using Stokes’ I
2. No correction vs. External Bias Temperature Calibration vs. OTT
Ext TB cal. partially diminishes the SSS misfit rms, OTT has a very
strong improvement effect.
Ext TB cal partially corrects for the land-sea transition effect and
seems to work better in the North Atl. waters (?)
The combined use of both techniques can be envisaged…
3. Model 2 vs. Model 3(16)
Model 3 is still in definition, conf. 16 (from WISE) has been used,
performing relatively close to Model 2. difference between models
are strongly related to SST.
4. All overpasses vs. Ascending vs. Descending
Waters appear fresher when land/ice enters in the FOV and saltier
when it exits when using only ascending or descending passes, the effect
is compensated using both. Descending passes give
generally saltier SSS.
Conclusions (+ or -)
Ext TB calibration gives more homogeneous results…again the
combined use can be envisaged…
5. EAF vs. AF
Using only AF FOV a positive bias has been found in the ascending
passes, negative in the descending, w.r.t the case of using EAF FOV.
Change in statistics is small.
6. July vs. August
August 10-day SSS misfit is in average 0.3-0.4 psu fresher than
July’s SSS misfit.
Anyway statistics are very similar and no clear improvement can
be observed.