Post on 28-Dec-2015
SMOS QWG-5, 30 May- 1 June 2011, ESRIN
Ocean Salinity
1
1. Commissioning reprocessing analysis
2. New processor version: improvements and problems detected/solved
3. Present performance
4. Future evolution: ongoing studies
SMOS QWG-5, 30 May – 1 June 2011, ESRIN
Land sea contamination correction
J. Martínez, V. González, C. Gabarró, J. Gourrion and BEC–TEAMSMOS Barcelona Expert CentrePg. Marítim de la Barceloneta 37-49, Barcelona SPAINE-mail: jfont@icm.csic.esURL: www.smos-bec.icm.csic.es
SMOS QWG-5, 30 May – 1 June 2011, ESRIN 3
Land contamination
Impact of correction implemented by Deimos on the strong halo around continental surfaces
to avoid multiplying the first Fourier parameter by the element of area (sqrt(3) * Distance_ratio * Distance_ratio/2)
L1PP run at BEC without and with correction
71 ascending orbits, 71 descending from 17-21 August 2010
Tb at 42.5º; filtering 40 < Tb < 200
Tb maps: average per ISEA GP and then average for 1º*r*cos(lat).
SSS semi-orbits (problem in running several orbits at a time)
SMOS QWG-5, 30 May – 1 June 2011, ESRIN
Tb ascending maps
SMOS QWG-5, 30 May – 1 June 2011, ESRIN
Tb descending maps
SMOS QWG-5, 30 May – 1 June 2011, ESRIN 6
Impact on SSS
SSS 3 semi-orbits
Run with patched L1PP and L2OS 3.17
Specific OTT computed from uncorrected and corrected L1
SMOS QWG-5, 30 May – 1 June 2011, ESRIN 7
Uncorrected
SMOS QWG-5, 30 May – 1 June 2011, ESRIN 8
Corrected
SMOS QWG-5, 30 May – 1 June 2011, ESRIN 9
Uncorrected
SMOS QWG-5, 30 May – 1 June 2011, ESRIN 10
Corrected
SMOS QWG-5, 30 May – 1 June 2011, ESRIN
Conclusion
The correction has removed the first order problem (strongest signal)
Back to the original scene dependant bias issue (A. Camps 2005)?
11
SMOS QWG-5, 30 May – 1 June 2011, ESRIN
Pre-launch semi-empirical roughness model (SSS3) was derived from data obtained during the WISE experiments (2000-2001) on an oil platform in the NW Mediterranean
New fitting using actual SMOS data (residual after removing the rest of modelled emission components)
Guimbard et al., “SMOS semi-empirical ocean forward model adjustment” submitted to TGRS SMOS special issue
12
New semi-empirical roughness model
SMOS QWG-5, 30 May – 1 June 2011, ESRIN
New semi-empirical roughness model
13
QWG-5, Frascati, May 30-31st, 2011
OTT sensitivity study
J. Gourrion, M. Portabella, R. Sabia, S.Guimbard
SMOS-BEC, ICM/CSIC
QWG-5, Frascati, May 30-31st, 2011
OTT sensitivity
DPGS OTT
Impact on OTT quality of different factors:
1. Number of snapshots used
2. Temporal variability and apparent drift
3. Latitudinal variability
Alternative OTT estimation strategy
Method and preliminary results
QWG-5, Frascati, May 30-31st, 2011
For a 16-days period dataset (Aug. 3rd – Aug 18th), about 12000 snapshots are available after comprehensive filtering (land, outliers, descending overpasses)
N OTTs are computed by randomly selecting n snapshots among all available. (N-1) rms difference of the N OTTs are then computed.
N decreases with increasing n, leading to N=2 when n=6000, i.e., about half of the total amount in the 16-days period.
For consistency, the same experiment is repeated for two additional 16-days periods (Aug. 19th – Sep 3rd, Sep. 4th – Sep 19th). The overall rms values are obtained by averaging the 3 16-day period scores.
As expected, OTT robustness depends on number of snapshots used. Current operational OTT has a 0.25K error only due to sampling.
OTT sensitivity
Impact of number of snapshots
QWG-5, Frascati, May 30-31st, 2011
OTT sensitivity
Temporal variability
A 48-days period dataset (August-Sept 2010) is used and split into 8-days subsets. Same filtering than previous experiment.
The reference situation is given by the first 8-days subset.
For each subset, a fixed number of snapshots are randomly selected to compute an OTT, n = 6250.
The OTT rms increase (relative to reference) indicates an increasing data inconsistency with time, i.e., apparent drift.
QWG-5, Frascati, May 30-31st, 2011
Ocean/ice transition
Salinity ? Rain ?
Roughness residuals ?New model 3SSA/SPM model
OTT sensitivity
Latitudinal variability
A 16-day period dataset (Aug. 3rd – Aug 18th) is used and split into 6° latitudinal band subsets.
The reference situation is given by the [36° S, 30° S] latitudinal band subset.
For each subset, a fixed number of snapshots are randomly selected to compute an OTT, n = 610.
The OTT rms differences (relative to reference) mainly indicate potential forward model and auxiliary data errors.
QWG-5, Frascati, May 30-31st, 2011
OTT sensitivity
OTT as mean departure from
full forward model: summary OTT robustness significantly depends on sampling. Current OTT
computation should use a larger number of snapshots.
Temporal inconsistencies due to non-modelled instrumental/reconstruction instability and imperfect Foreign Sources modelling
Latitudinal inconsistencies due to imperfect modelling or auxiliary parameters
OTTs estimated from different datasets will vary depending on the distribution of sampled geophysical conditions
With current OTT methodology, the data are adjusted to reproduce the mean forward model behaviour (e.g., angular dependency): updated forward models are NOT independent from pre-launch versions (used to compute the OTT)
QWG-5, Frascati, May 30-31st, 2011
OTT sensitivity
Objective:
Estimate systematic errors in the antenna frame
while avoiding use of forward models as much as possible
Main differences with current OTT:
do not use forward models do not assume that geophysical variability is negligible
BUTselect specific environmental conditions (U,SST,SSS,low galactic,…)
MEAN angular dependency is fitted with a simple polynomial function and removed from the mean scene to obtain the systematic error pattern
Work in progress: only five days of data processed in this study.
New OTT estimation method: basics (1)
QWG-5, Frascati, May 30-31st, 2011
OTT sensitivity
New OTT estimation method: comparison
INCONSISTENT ANGULAR DEPENDENCE BETWEEN SMOS DATA AND
PRE-LAUNCH FORWARD MODELS
QWG-5, Frascati, May 30-31st, 2011
OTT sensitivity
New OTT estimation method: stability (1)
Selecting different wind speed conditions
RMS VALUES CONSISTENT WITH EXPECTED VALUES FROM NUMBER OF SAMPLES – GRANULAR PATTERNS
QWG-5, Frascati, May 30-31st, 2011
OTT sensitivity
New OTT estimation method: summary
Adequate data selection techniques + mean angular dependence removal allows to obtain ROBUST OTT estimates WITHOUT introducing systematic errors from imperfect forward model and auxiliary information
Temporal drift effects still need to be accounted for.
Angular dependence of the corrected images is consistent with the original SMOS data
Work in progress:
Use more data Further analyze latitudinal and temporal variations New GMF fit using new OTT
Near-future work will compare the goodness of either additive or multiplicative formulations.