SMOS Science Workshop, Arles, th Sept, 2011 IMPROVING SMOS SALINITY RETRIEVAL: SYSTEMATIC ERROR DIAGNOSTIC J. Gourrion, R. Sabia, M. Portabella, S. Guimbard, J. Tenerelli SMOS-BEC, ICM/CSIC CLS
SMOS Science Workshop, Arles, th Sept, 2011 Introduction Systematic errors in the SMOS reconstructed brightness temperature images identified rapidly after launch (J.Tenerelli) Data from March, 2010 X-pol Y-pol ξ ξ ηη
SMOS Science Workshop, Arles, th Sept, 2011 (*) AGP: antenna gain pattern Image reconstruction non-identical AGP (*) imperfectly known AGP (*) Imperfect calibration Error correction Foreign sources removal Measured visibilities Level 0 Calibrated visibilities Level 1ALevel 1B SMOS T B Introduction Systematic T B errors: why ? … as anticipated by Camps [1998], Anterrieu [2003]
SMOS Science Workshop, Arles, th Sept, 2011 Ocean Target Tranformation 1 st approach: overall systematic error correction in the antenna frame To avoid systematic inconsistencies between data and model during inversion, this fully empirical approach is convenient to optimize the retrieved salinity fields for a given instrumental and modeling state. This approach is operationally used in the L2OS processor
SMOS Science Workshop, Arles, th Sept, 2011 Ocean Target Tranformation 1 st approach: overall systematic error correction in the antenna frame Number of scenes Temporal variability Latitudinal variability from Gourrion et al. 2011, submitted to GRSL DPGS data from August 2010, Ascending passes
SMOS Science Workshop, Arles, th Sept, 2011 Ocean Target Tranformation 1 st approach: overall systematic error correction in the antenna frame Summary The overall error pattern has 2 components: Azimuthally-distributed systematic errors likely due to antenna patterns Incidence angle-dependent systematic errors data ? model ? The estimated pattern is highly variable with the dataset used to compute it Inconsistent with “systematic” errors
SMOS Science Workshop, Arles, th Sept, 2011 Ocean Target Tranformation 2 nd approach: specific error correction Characterize systematic errors in the antenna frame independently of forward models Get a stable estimate of the systematic error pattern Separate azimuthally distributed errors (antenna pattern- related) from other errors (data or model). Mandatory for consistent model improvement tasks and combination of measurements at same incidence but different location in the image Objectives
SMOS Science Workshop, Arles, th Sept, 2011 Ocean Target Tranformation 2 nd approach: specific error correction Requirements Homogeneous dataset in terms of environmental conditions Geophysical surface conditions data selection (U, SSS, SST) Ionospheric effects correction (aux TEC, T3) Extra-terrestrial contributions Sun : always present Level 1 sun correction activated Reflected Sky : data selection Stability focus on T1
SMOS Science Workshop, Arles, th Sept, 2011 Ocean Target Tranformation 2 nd approach: specific error correction Select specific geophysical conditions (U, SST, SSS) at individual (xi,eta) points using thresholds on auxiliary parameters Methodology Wind speed : U = U0 ± 0.5 m/s Sea surface salinity and temperature such that dielectric properties are nearly homogeneous: T b flat = ± 0.5 * ΔU
SMOS Science Workshop, Arles, th Sept, 2011 Ocean Target Tranformation 2 nd approach: specific error correction Select specific geophysical conditions (U, SST, SSS) at individual (xi,eta) points using thresholds on auxiliary parameters Methodology Sky reflections Courtesy of J. Tenerelli
SMOS Science Workshop, Arles, th Sept, 2011 Ocean Target Tranformation 2 nd approach: specific error correction Select specific geophysical conditions (U, SST, SSS) at individual (xi,eta) points using thresholds on auxiliary parameters Methodology Rotate polarization frame from antenna (X/Y) to surface (H/V), geometry+Faraday From the mean scene, fit its incidence angle (θ) dependence to obtain a simplified one-parameter empirical model Average T B H/V (ξ,η) – T B model (θ) in the antenna frame Rotate back from surface to antenna polarization frame (geometry) Sky reflections
SMOS Science Workshop, Arles, th Sept, 2011 Ocean Target Tranformation 2 nd approach: specific error correction Robustness (1): varying wind speed X-pol Y-pol 6 m/s – 8 m/s10 m/s – 8 m/s12 m/s – 8 m/s 10 o S > lat > 30 o S Between 5 and 11 m/s, pattern discrepancy is lower than 0.1 K r.m.s. |U-U 0 | < 1 m/s 16-days datasets
SMOS Science Workshop, Arles, th Sept, 2011 Ocean Target Tranformation 2 nd approach: specific error correction Robustness (1): varying wind speed X-pol Y-pol 6 m/s – 8 m/s10 m/s – 8 m/s12 m/s – 8 m/s 10 o S > lat > 30 o S Between 5 and 11 m/s, pattern discrepancy is lower than 0.1 K r.m.s. |U-U 0 | < 1 m/s Between 5 and 11 m/s, pattern discrepancy is lower than 0.1 K r.m.s.
SMOS Science Workshop, Arles, th Sept, 2011 Ocean Target Tranformation 2 nd approach: specific error correction Robustness (2): varying latitude range [35 o S,10 o S]-[55 o S,35 o S] X-pol Y-pol 6 m/s 8 m/s10 m/s12 m/s Strong discrepancy between different latitude bands due to varying sun alias location and imperfect sun removal procedure
SMOS Science Workshop, Arles, th Sept, 2011 Ocean Target Tranformation 2 nd approach: specific error correction Robustness (2): varying latitude range [35 o S,10 o S]-[55 o S,35 o S] X-pol Y-pol 6 m/s 8 m/s10 m/s12 m/s Depending on sun alias location, strong discrepancy between different latitude bands may appear due to imperfect sun tails removal procedure Nov 2010Aug 2010
SMOS Science Workshop, Arles, th Sept, 2011 Ocean Target Tranformation 2 nd approach: specific error correction Robustness (3): ascending / descending X-polY-pol [35 o S,10 o S], 8 m/s Faraday rotation is poorly accounted using the auxiliary TEC information. Ascending and descending passes cannot be combined together. 1 st Stokes is affected by galactic contamination in descending passes 1 st Stokes
SMOS Science Workshop, Arles, th Sept, 2011 Ocean Target Tranformation 2 nd approach: specific error correction Robustness (4): varying temporal window The number of observations used in estimating the error pattern is crucial regarding its robustness Patterns obtained over different but consistent geophysical conditions can be combined to further increase the robustness 1 st Stokes 6 m/s 8 m/s 10 m/s 12 m/s
SMOS Science Workshop, Arles, th Sept, 2011 Ocean Target Tranformation 2 nd approach: specific error correction Robustness (4): varying temporal window X- and Y-pol patterns are contaminated by a rotation-related pattern Y-pol X-pol 1 st Stokes Patterns determined over different time periods cannot be safely averaged together
SMOS Science Workshop, Arles, th Sept, 2011 Ocean Target Tranformation 2 nd approach: specific error correction Robustness (5): varying wind speed interval width X-pol ΔU = 0.7 m/s 6 m/s – 8 m/s10 m/s – 8 m/s12 m/s – 8 m/s ΔU = 1 m/s ΔU = 2 m/s The number of observations used in estimating the error pattern is crucial regarding its robustness Patterns obtained over different but consistent geophysical conditions can be combined to further increase the robustness
SMOS Science Workshop, Arles, th Sept, 2011 Ocean Target Tranformation Difference between both approaches X-polY-pol 30 o 50 o 40 o 20 o 50 o 40 o 30 o 20 o The 1 st approach OTT includes systematic discrepancy with incidence angle between data and models which origin are presently not identified.
SMOS Science Workshop, Arles, th Sept, 2011 Summary The present OTT as implemented at DPGS is dependent on imperfect forward models variable from one dataset to the other (~0.5 K) contaminated by residual sun correction errors (~0.5 K near sun tails) An alternative method to estimate systematic error patterns is proposed Galactic contribution intensity drives the choice of the dataset Stable over various geophysical conditions (~0.1 K for 5 < U < 11 m/s) Importance of data selection Difficulty to mix ascending/descending passes (Faraday, Galactic) Further work: compare with other low-galactic datasets (A/D), Faraday from T3
Next plenary workshop foreseen in March 2012 Additional institutions and countries are welcome! SMOS-Mission Oceanographic Data Exploitation SMOS-MODE SMOS-MODE supports the network of SMOS ocean-related R&D
SMOS Science Workshop, Arles, th Sept, 2011 SMOS-MODE – SMOS-Mission Oceanographic Data Exploitation SMOS-MODE supports the network of SMOS ocean-related R&D Meetings Workshops Training school Short term scientific missions Overall Aim: To coordinate pan-European teams to define common protocols to produce high-level salinity maps and related products, and broaden expertise in their use for operational applications. To bridge remote sensing and applications communities 14 countries represented so far. Co-chairs: Antonio Turiel, SMOS Barcelona Expert Centre (SMOS-BEC), Barcelona, Spain Nicolas Reul, IFREMER, Brest, France Next plenary workshop foreseen in March 2012 Additional institutions and countries are welcome!