model-independent estimation of systematic errors in smos brightness temperature images
DESCRIPTION
MODEL-INDEPENDENT ESTIMATION OF SYSTEMATIC ERRORS IN SMOS BRIGHTNESS TEMPERATURE IMAGES J. Gourrion, S. Guimbard, R. Sabia, M. Portabella, V. Gonzalez, A. Turiel, J. Ballabrera, C. Gabarró, F. Perez, J. Martinez SMOS-BEC, ICM/CSIC [email protected]. Introduction. - PowerPoint PPT PresentationTRANSCRIPT
SPCM-9, Esac, May 3rd, 2012
MODEL-INDEPENDENT ESTIMATION OF SYSTEMATIC ERRORS IN SMOS BRIGHTNESS
TEMPERATURE IMAGES
J. Gourrion, S. Guimbard, R. Sabia, M. Portabella, V. Gonzalez, A. Turiel, J. Ballabrera, C. Gabarró, F. Perez, J. Martinez
SMOS-BEC, ICM/CSIC
SPCM-9, Esac, May 3rd, 2012
Introduction
Forwardmodel
Auxiliarydata
Level 2
Retrieved SSS
Retrieval scheme
Reconstructed TBs
Level 1B/C
Corrected TBs (OTT)
Optimal Salinity retrieval for given dataset and given forward model
Adjust measurements to model on average
Reduce overall SSS biases
Forwardmodel
Auxiliarydata
Y-pol
ξ
η
SPCM-9, Esac, May 3rd, 2012
Introduction
SMOS-retrieved SSS biases due to forward model imperfections at high wind speed
from Guimbard et al. 2012, TGRS
Forward model errors (roughness, galactic, Faraday, …) contribute to a variability of the estimated pattern of about 0.5 K
SPCM-9, Esac, May 3rd, 2012
OTT - Current approach
Number of scenes
Temporal variability
Latitudinal variability
from Gourrion et al. 2012, GRSLDPGS data from August 2010,Ascending passes
The estimated pattern varies with the dataset used – typically 0.5 KThis includes the variability of model errors.
Overall misfit between data and model: stability
SPCM-9, Esac, May 3rd, 2012
OTT uncertainty: 0.5 K
Introduction
Forwardmodel
Auxiliarydata
Level 2
Retrieved SSS
Retrieval scheme
Reconstructed TBs
Level 1B/C
Further salinity improvement requires forward model improvement
Need for a model independent correction
Might be valid forOcean/Ice/Land images
Corrected TBs (OTT)
SPCM-9, Esac, May 3rd, 2012
• Characterize systematic errors in the antenna frame independently of forward models –
mandatory for consistent model improvement tasks
• Get a stable estimate of the systematic error pattern variability tipically lower than 0.5 K
Objectives
OTT - New approach
Our ocean results are compared with those obtained by F.Cabot using SMOS data acquired over ice at Dome-C
SPCM-9, Esac, May 3rd, 2012
Strategy (Ocean – Ice)
• Use a dataset with low geophysical/environmental variability
data selection (U,SSS,SST,galaxy) – stable target, single point at Dome-C
OTT - New approach
June 2010
Dec. 2010
June 2011
SPCM-9, Esac, May 3rd, 2012
Strategy (Ocean – Ice)
• Use a dataset with low geophysical/environmental variability
• Rotate from antenna (X/Y) polarization frame to surface (H/V) - geometry+Faraday
OTT - New approach
SPCM-9, Esac, May 3rd, 2012
H-pol TB H-pol TB
Strategy (Ocean – Ice)
• Use a dataset with low geophysical/environmental variability
• Rotate polarization frame from antenna (X/Y) to surface (H/V) - geo+Faraday
• From the mean scene, fit its incidence angle (θ) dependence to obtain a simplified one-parameter empirical model – H/V
OTT - New approach
SPCM-9, Esac, May 3rd, 2012
Strategy (Ocean – Ice)
• Use a dataset with low geophysical/environmental variability
• Rotate polarization frame from antenna (X/Y) to surface (H/V) - geo+Faraday
• From the mean scene, fit its incidence angle (θ) dependence to obtain a simplified one-parameter empirical model – H/V
• Rotate back to get the expected X/Y TBs for all selected data
OTT - New approach
SPCM-9, Esac, May 3rd, 2012
Strategy (Ocean – Ice)
• Use a dataset with low geophysical/environmental variability
• Rotate polarization frame from antenna (X/Y) to surface (H/V) - geo+Faraday
• From the mean scene, fit its incidence angle (θ) dependence to obtain a simplified one-parameter empirical model – H/V and get the anomaly
• Rotate back to get the expected X/Y TBs for all selected data
• Compute the anomaly, mean difference between data and model
OTT - New approach
X-pol TB anomaly Y-pol TB anomaly
SPCM-9, Esac, May 3rd, 2012
6 m/s – 8 m/s 10 m/s – 8 m/s 12 m/s – 8 m/s
June 2010
6 m/s – 8 m/s 10 m/s – 8 m/s 12 m/s – 8 m/s
December 2010
6 m/s – 8 m/s 10 m/s – 8 m/s 12 m/s – 8 m/s
June 2011
6 m/s – 8 m/s 10 m/s – 8 m/s 12 m/s – 8 m/s
December 2011
6 m/s 8 m/s 10 m/s 12 m/s
June 2010
Robustness (1): varying wind speed
(XX+YY)/2
Between 5 and 11 m/s, pattern discrepancy is lower than 0.05 K r.m.s.
|U-U0| < 1 m/s
18-days datasets
OTT - New approach
SPCM-9, Esac, May 3rd, 2012
Robustness (2): varying time period
RMS differences over 1 year interval lower than 0.15 K Related to residual calibration errors or instrument stability ?
(XX+YY)/2
OTT - New approach
Same latitudinal band
Same season
Same celestial reflections
Same sun location
[55oS, 35oS]
June 2011 - 2010
[35oS, 0oS]
December 2011 - 2010
January 2012 - 2011
[35oS, 0oS]
SPCM-9, Esac, May 3rd, 2012
Robustness (3): comparing Ocean/Ice results
Ocean
OTT - New approach
Y-pol
X-pol
Ice
Results over ice provided by F.Cabot
SPCM-9, Esac, May 3rd, 2012
Robustness (3): comparing Ocean/Ice resultsOcean
OTT - New approach
Y-pol
X-polIce
from F.CabotIce
with Ocean method Ice
with modified Ocean method
We can define a method so that differences in Ocean/Ice results are not methodological
SPCM-9, Esac, May 3rd, 2012
Robustness (3): comparing Ocean/Ice results
High consistency between Ocean-derived and Ice-derived systematic error patterns
Residual differences to be understood. Reconstruction errors ? Ongoing work …
Ocean
OTT - New approach
Y-pol
X-pol
Ice
SPCM-9, Esac, May 3rd, 2012
Summary
Near-future improvement in SMOS salinity products will come with forward model adjustment (roughness, Faraday, galactic reflection, …)
Model improvement tasks require a specific approach for systematic error correction
Model-independent
Stability lower than 0.5 K
The approach proposed, apart from being model-independent, is
stable when estimated from datasets with different geophysical conditions (< 0.1 K r.m.s)
stable over time, in the limit of instrument stability (< 0.15 K)
promising consistency with independent results obtained over ice surfaces at Dome-C (F.Cabot) the pattern is robust
SPCM-9, Esac, May 3rd, 2012
Summary
Further work:
Investigate origin of residual Ocean/Ice inconsistencies (inc. angle)
Forward model improvement
Revisit roughness contribution
Faraday rotation: ongoing work