Download - EXTENDING THE LAND SEA CONTAMINATION CHARACTERIZATION TO THE EXTENDED ALIAS-FREE FIELD OF VIEW
EXTENDING THE LAND SEA CONTAMINATION CHARACTERIZATION TO THE EXTENDED ALIAS-
FREE FIELD OF VIEW
Joe Tenerelli (CLS) and Nicolas Reul (IFREMER)SMOS Quality Working Group #15
ESA ESRIN6-8 October 2014
REMINDER OF THE PROBLEM
There is a significant bias in the retrieved surface salinity around all of the continents, and this bias seems to be stable over the entire mission. By significant I mean that the biases can exceed 2 pss, while the range of SSS over the global oceans is 30-40 pss (away from river plumes and freshwater lenses).
We will call this bias land-sea contamination, or LSC.
AN EXAMPLE OF LSC AROUND AUSTRALIA
BASED UPON ARGO MEASUREMENTS FROM 2010-2013
AN EXAMPLE OF LSC AROUND SOUTH AMERICA
BASED UPON ARGO MEASUREMENTS FROM 2010-2013
GLOBAL LSC IN TERMS OF FIRST STOKES PARAMETER
ASCENDING PASSES FOR MAY 2011: FIRST STOKES PARAMETER BIAS
The LSC is global:
ASCENDING PASSES FOR MAY 2011: RETIEVED SALINITY BIAS
GLOBAL LSC IN TERMS OF RETRIEVED SALINITYThe LSC is global:
ASCENDING PASSES FOR MAY 2011: RETIEVED SALINITY BIAS
Note: In this presentation SMOS SSS is retrieved using the first Stokes parameter and a simple linear retrieval algorithm.
GLOBAL LSC IN TERMS OF RETRIEVED SALINITY
LSC IS SCENE-DEPENDENTThe biases are also scene-dependent and therefore change as the distribution of land over the front half space changes:
THE LSC CORRECTION
Given the preceding, we suppose that the ‘land contamination’ bias in polarization p may be expressed as
where is the pass direction, are the geographic longitude and latitude and are the usual director cosine coordinates.
As a first test we developed a coarse lookup table over just the alias-free field of view (AF-FOV). Here we extend the results to the extended alias-free field of view (EAF-FOV).
∆Tp(D,λg,θg,ξ,η),
KEY POINTS OF THE EMPIRICAL LSC CORRECTION
• latitude-longitude grid: 0.5x0.5 deg lat-lon grid
• Director cosine grid: 0.025x0.025 director cosine units rather than 0.1x0.1 director cosine units
• Two lookup tables are calculated: One for RFI-flagged snapshots and another for non-RFI-flagged (nominal) snapshots.
• 41,469 half-orbits are used to compute the mission average land contamination.
• Biases in all four Stokes parameters are computed.
• Final correction is weighted by a function that approaches zero as the fraction of land in the front half space approaches 0.2% and approaches one as the fraction approaches 1%.
• The correction also approaches zero as the ice fraction approached 0.02%.
DISCRETIZATION
PROCESSING STEPS (1)
Successive snapshots (‘scene’) (Tx,Ty,Uxy,Vxy) on hex grid
Flag scene for RFI:Tx > 500 K orTy > 500 K orUxy > 200 K
Compute bias relative to forward model (ref. SSS is WOA-2009) for all
four Stokes parameters in instrument polarization basis
Grid biases onto ISEA-4H9 grid and identify (xi,eta) bins
Accumulate data into daily maps on 0.5°x0.5° lat-lon grid; remove
gridpoints with no measurements; introduce linear indexing for
retained gridpoints
Accumulate data into monthly maps; keep track
of pointwise measurement counts
Create separate monthly files for each variable
Average the monthly Stokes parameter bias
maps over the full period (Jan 2010-June 2014)
RFI-flagged RFI-free
PROCESSING STEPS (2)
Remove gridpoints with fewer than 10 passes entering into the
average
For each (xi,eta) cell, remove average bias over all (lat,lon) cells with no land or ice in the front
half-space; removes impact of the choice of
OTT
Apply linear weighting function to biases which
ramps down from 1 at FHS land fraction of 1% to 0 at
land fraction of 0.2%.
Apply similar weighting function that ramps to 0 as the ice fraction
approaches 0.02%.
Set correction to zero south of 60°S latitude
Merge all Stokes parameter biases (single
precision) into a single file for each pass direction
and for RFI and non-RFI-flagged scenes.
Each file is 2.45 GB in size with about 88 million
gridpoints in each.
DISCRETIZATION
EXAMPLE BIAS OVER EAF-FOV
MEASUREMENT COUNT FOR THE AVERAGINGnon-RFI-flagged snapshots
EXAMPLE MAP OF THE FIRST STOKES BIASnon-RFI-flagged snapshots
MEASUREMENT COUNTS FOR THE AVERAGINGRFI-flagged snapshots
EXAMPLE MAP OF THE FIRST STOKES BIASRFI-flagged snapshots
EVALUATION OF THE CORRECTION
1. Global perspective for May 2011
2. Impact in areas with stable SSS: Australia and South America in May 2011 using ISAS as the reference
3. Comparison with ARGO data around Australia and South America over the period 2010-2013
4. Impact in areas with strongly varying SSS: Panama and the Amazon plume regions
EVALUATION OF THE CORRECTION
1. Global perspective for May 2011
2. Impact in areas with stable SSS: Australia and South America in May 2011 using ISAS as the reference
3. Comparison with ARGO data around Australia and South America over the period 2010-2013
4. Impact in areas with strongly varying SSS: Panama and the Amazon plume regions
INITIAL EVALUATION OF THE CORRECTION USING THE MONTH OF MAY 2011 AND THE ISAS MONTHLY MAP
FIRST STOKES PARAMETER DIVIDED BY TWO BEFORE CORRECTION
INITIAL EVALUATION OF THE CORRECTION USING THE MONTH OF MAY 2011 AND THE ISAS MONTHLY MAP
FIRST STOKES PARAMETER DIVIDED BY TWO AFTER CORRECTION
INITIAL EVALUATION OF THE CORRECTION USING THE MONTH OF MAY 2011 AND THE ISAS MONTHLY MAP
DESCENDING-ASCENDING BIAS BEFORE CORRECTION
INITIAL EVALUATION OF THE CORRECTION USING THE MONTH OF MAY 2011 AND THE ISAS MONTHLY MAP
DESCENDING-ASCENDING BIAS AFTER CORRECTION
EVALUATION OF THE CORRECTION
1. Global perspective for May 2011
2. Impact in areas with stable SSS: Australia and South America in May 2011 using ISAS as the reference
3. Comparison with ARGO data around Australia and South America over the period 2010-2013
4. Impact in areas with strongly varying SSS: Panama and the Amazon plume regions
INITIAL EVALUATION OF THE CORRECTION USING THE MONTH OF MAY 2011 AND THE ISAS MAPS
First consider two areas with stable surface salinity whose distribution is well-measured by ARGO floats.
INITIAL EVALUATION OF THE CORRECTION USING THE MONTH OF MAY 2011 AND THE ISAS MAPS
First consider two areas with stable surface salinity whose distribution is well-measured by ARGO floats.
INITIAL EVALUATION OF THE CORRECTION USING THE MONTH OF MAY 2011 AND THE ISAS MAPS
First consider two areas with stable surface salinity whose distribution is well-measured by ARGO floats.
INITIAL EVALUATION OF THE CORRECTION USING THE MONTH OF MAY 2011 AND THE ISAS MAPS
INITIAL EVALUATION OF THE CORRECTION USING THE MONTH OF MAY 2011 AND THE ISAS MAPS
CORRECTION REDUCES VARIATION OF BIAS WITH LAND FRACTION OUTSIDE THE FUNDAMENTAL HEXAGON
INITIAL EVALUATION OF THE CORRECTION USING THE MONTH OF MAY 2011 AND THE ISAS MAPS
CORRECTION REDUCES VARIATION OF BIAS WITH LAND FRACTION OUTSIDE THE FUNDAMENTAL HEXAGON
INITIAL EVALUATION OF THE CORRECTION USING THE MONTH OF MAY 2011 AND THE ISAS MAPS
INITIAL EVALUATION OF THE CORRECTION USING THE MONTH OF MAY 2011 AND THE ISAS MAPS
BIAS APPEARS AS SOON AS LAND APPEARS OUTSIDE THE FUNDAMENTAL HEXAGON
INITIAL EVALUATION OF THE CORRECTION USING THE MONTH OF MAY 2011 AND THE ISAS MAPS
BIAS APPROACHES 2.5 PSU AS THE LAND FRACTION APPROACHES 0.9.
INITIAL EVALUATION OF THE CORRECTION USING THE MONTH OF MAY 2011 AND THE ISAS MAPS
INITIAL EVALUATION OF THE CORRECTION USING THE MONTH OF MAY 2011 AND THE ISAS MAPS
INITIAL EVALUATION OF THE CORRECTION USING THE MONTH OF MAY 2011 AND THE ISAS MAPS
CORRECTION SEEMS TO REDUCE THE ABSOLUTE ERROR TO ABOUT 150-200 KM FROM THE COAST. CORRECTION IS LESS EFFECTIVE WITHIN 150 KM OF THE COAST.
EVALUATION OF THE CORRECTION
1. Global perspective for May 2011
2. Impact in areas with stable SSS: Australia and South America in May 2011 using ISAS as the reference
3. Comparison with ARGO data around Australia and South America over the period 2010-2013
4. Impact in areas with strongly varying SSS: Panama and the Amazon plume regions
ARGO COLLOCATIONS: AUSTRALIA
ARGO COLLOCATIONS: AUSTRALIA
TIME PERIOD: 2010-2013
ARGO-SMOS collocations are binned into 50 km wide bands (as a function of distance to coast). Biases are then computed in these bands.
ARGO COLLOCATIONS: AUSTRALIA
UNCORRECTED SMOS SSS
ARGO COLLOCATIONS: AUSTRALIA
CORRECTED SMOS SSS
ARGO COLLOCATIONS: AUSTRALIA
ARGO COLLOCATIONS: AUSTRALIA
bias reduction of about 0.5 pss
ARGO COLLOCATIONS: AUSTRALIA
Impact of the LSC correction drops rapidly within about
200 km of the coast
ARGO COLLOCATIONS: AUSTRALIA
bias reduction of about 0.8 pss
A COMPARISON WITH SHIP TSG DATA
WITHOUT LSC CORRECTION
A COMPARISON WITH SHIP TSG DATA
WITH LSC CORRECTION
ARGO COLLOCATIONS: SOUTH AMERICA
ARGO COLLOCATIONS: SOUTH AMERICA
ARGO COLLOCATIONS: SOUTH AMERICA
ARGO COLLOCATIONS: SOUTH AMERICA
ARGO COLLOCATIONS: SOUTH AMERICA
bias reduction of about 1 pss
ARGO COLLOCATIONS: SOUTH AMERICA
bias reduction of about 2 pss
EVALUATION OF THE CORRECTION
1. Global perspective for May 2011
2. Impact in areas with stable SSS: Australia and South America in May 2011 using ISAS as the reference
3. Comparison with ARGO data around Australia and South America over the period 2010-2013
4. Impact in areas with strongly varying SSS: Panama and the Amazon plume regions
DYNAMIC ZONE: PANAMAComparing Matisse transect with SMOS SSS
DYNAMIC ZONE: PANAMAComparing Matisse transect with SMOS SSS
DYNAMIC ZONE: PANAMA
bias reduction of nearly 2 pss
DYNAMIC ZONE: PANAMAComparing Matisse transect with SMOS SSS
DYNAMIC ZONE: PANAMAComparing Matisse transect with SMOS SSS
DYNAMIC ZONE: PANAMA
bias reduction of about 1 pss
AMAZON PLUME: ANACONDA TSGComparing Anaconda transects with 10-day SMOS SSS
AMAZON PLUME: ANACONDA TSGComparing Anaconda transects with 10-day SMOS SSS
AMAZON PLUME: ANACONDA TSGComparing Anaconda transects with 10-day SMOS SSS
AMAZON PLUME: ANACONDA TSGComparing Anaconda transects with 10-day SMOS SSS
Corrected SMOS SSS differs from the reference used to derive the LSC correction.
AMAZON PLUME: ANACONDA TSG
AMAZON PLUME: ANACONDA TSG
AMAZON PLUME: ANACONDA TSG
AMAZON PLUME: ANACONDA TSG
AMAZON PLUME: ANACONDA TSG
AMAZON PLUME: ANACONDA TSG
AMAZON PLUME: ANACONDA TSG
AMAZON PLUME: ANACONDA TSG
AMAZON PLUME: ANACONDA TSGLSC correction seems to overcorrect the bias (when compared to the ship TSG) in two zones along the ship track. In fact, the corrected SSS exceeds the reference used to derive the LSC correction (black curve)! To be investigated further…
CONCLUSIONS
• Introduced the new land sea contamination lookup table, now defined over the extended alias-free field of view.
• Comparison with ISAS monthly SSS map for May 2011 around Australia and South America shows reduction in LSC. Comparisons with ARGO from 2010-2013 also show reduction in the LSC.
• Reductions in LSC may also be seen in regions of strongly time varying SSS such as near Panama and in the Amazon plume region. However, results are mixed in these areas and cases have been found in which application of the correction seems to increase the biases relative to in-situ measurements. This must be investigated further.
REMAINING ISSUES• Difficult to derive the correction in RFI-contaminated areas such as the west Pacific, north
Atlantic and Pacific, and some areas around Africa and South America.
• Impossible to derive the correction where the ice distribution varies with time in the field of view.
• Correction lookup table remains contaminated in some areas, possible due to residual RFI that is not detected by the snapshot filter.
• Latitudinal drift is not corrected when deriving the correction, and this drift may impact the correction.
• Method of handling RFI requires testing and refinement.
• A forward model is required to derive the correction, and errors in this model will impact the correction.
• Reference for salinity for the forward model is WOA-2009. Need to examine use of better reference such as Aquarius SSS maps or ISAS maps.
• Analysis based upon v500 data. Analysis needs to be redone using the data from the latest reprocessing campaign.
AMAZON PLUME
AMAZON PLUME
AMAZON PLUME
AMAZON PLUME