Transcript OS E2E STUDY C. Mugerin – ARGANS LTD

OS E2E STUDY
C. Mugerin – ARGANS LTD
SMOS AlgoVal #14
Barcelona, 29-30 Jan. 2009
Outlines
•
•
•
•
•
•
•
•
MIRE case study
Algorithm validation plan
Software environment
Perfect TB
L1OP reconstruction
SEPS geometry issues
L1OP reconstructed TB with SEPS rotation
Conclusion
MIRE case study
MIRE case study
Algorithm validation plan
•
•
•
•
•
SSS retrieval
Front effect
Coast effect
Ice effect
Cardioid
Software environment
•
SEPS v1.6
–
–
–
–
–
•
Updated scenarios directory: removed
obsolete scenarios, new FTT scenario
generated with a different orbit file;
DPGS L0 data header have been updated:
correlator layer and SEPS-GS version;
Back-lobes correction due to respect L1PP
version 1.6 definitions, to mitigate SPR
389;
Error Budget activities:
• SEPS-GS ‘Error’ and ‘No Error’ modes
improved, SPR 384;
• New SRN section on this subject;
Corrected the following SPR:
• SPR 290 Only under windows and
using Matlab 6.5 is the ionospheric
correction still being simulated by
the Iri95 routine;
L1PP 1.6.0
–
–
Flat target transformation
Gibbs 1 error correction
Perfect TB
SSS
Perfect TB
SSS error
Perfect TB
SSS
Perfect TB
SSS error
Perfect TB
SST
Perfect TB
UN10
Perfect TB
VN10
L1OP reconstruction
SSS
L1OP reconstruction
SSS error
L1OP reconstruction
SSS
L1OP reconstruction
SSS error
SEPS GEOMETRY ISSUES
SEPS geometric rotation
SEPS geo-localization
Extracted from SO-DS-DME-L1PP-0009
L1OP RECONSTRUCTED TB WITH SEPS ROTATION
SEPS rotation
SSS
SEPS rotation
SSS error
SEPS rotation
SSS
SEPS rotation
SSS error
SEPS rotation
SSS
Retrieval with L1 rotation
Retrieval with SEPS rotation
SEPS rotation
SSS residuals maps
SEPS rotation
SSS residuals histograms
SEPS rotation
SSS retrieval cross-plot
SEPS rotation
AlgoVal #13
Extracted from L1PP TB corrections strategies
November 20, 2008
Conclusion
•
•
•
•
•
All the blame was wrongly put on the bias in TB reconstruction. A large improvement resulted from
the use the geometrical angles coming from SEPS. Now that the “correct” angles are used in the SSS
retrieval, what is left as for the SSS bias is strongly believed to be coming from TB reconstruction
bias. The good news is that the bad areas are perfectly correlated now with the shift in the azimuth:
the worse the shift the worse the retrieved SSS bias. These are larger error around the poles and
extremely small at the equator relative to the error map from the perfect instrument.
Additionally the VB map deformation if it is confirmed might take part in the explanation of the
change of bias on TB depending on proportion of land. A quick test would be to generate a mat file
with the angles coming from L1 and them instead of calling ORINC in SEPS-ACRI to load directly
these data so that SEPS can proceed with the correct angles.
If Gibbs 2 algorithm is called in L1 processor which makes use of the geometrical angle more error
might be introduced than corrected. The residual TB map that will be processed might contain in
the end more heterogeneities than the original one…
Based on those elements, the calculation of the geometrical angle in SEPS should be corrected if it
is not done in version 2.0. Furthermore it seems that in the end the geometric rotation has to be a
CFI itself. It is a purely geometrical concept which has little to do with the underlying science.
Therefore a CFI would allow every module to speak the same language.
In those circumstances the full potential of L2OS algorithm cannot be assessed. Though time might
be too short to consider a full upgrade of all components of the end-to-end chain, it should be
stressed out that without an accurate simulator that can accurately predict what can be expected
from the science implemented in the processors, it would be tedious to decide whether or not the
actual accuracy coming out of the real instrument are optimal.