1. LEO Calibration over Rayleigh Scattering …toward the ATBD
GSICS Annual Meeting 4-8th March 2013, Williamsburg, VA
LEO Calibration over Rayleigh Scattering
…toward the ATBD
Bertrand Fougnie
Patrice Henry
Contributions Lydwine Gross-Colzy and Elise Durant (Cap Gemini, France)

2. LEO Calibration over Rayleigh Scattering
… toward the ATBD
Context
Cross-calibration LEO/GEO over desert sites
Outlines of the ATBD
Selection of scenes
Selection of clear pixels
Computation of the TOA signal
Calibration and matchups
A quick look on the error budget
Examples of applications
Why not immediately applicable to GEO

3. Context
Calibration over Rayleigh scattering appears to be a powerful method to :
validate in-flight calibration if an on-board device exists
contribute to elaborate the in-flight calibration if no onboard device
in addition to other methods using natural targets (“vicarious”)
validate the monitoring or temporal drift
validate the calibration inside the field of view
analyze instrumental behavior (straylight, polarization , spectral…)
This method was developed for LEO sensors and intensively used
Calibrate various sensors over this “identical” reference provide potentially
not a direct cross-calibration, but a calibration over the same reference
Nevertheless this method is still not fully mature for GEO
An ATBD has to be written
A first version based on the LEO experience
A second version including extension to GEO

4. Method Statistical approach over molecular scattering (Rayleigh) :
observe the atmosphere above ocean surface (= dark surface)
calibration from blue to red 443nm to 670nm
contributions to the TOA signal
Rayleigh molecular scattering : accurately computed (SOS code)
main contributor : ~85/90% of the TOA signal
ocean surface : prediction through a climatology
no foam because of threshold on wind speed
aerosols : rejected using threshold + corrected
background residue using 865nm band + Maritime-98 model
for POLDER (and in general) criteria : <ta> = and max(ta) = 0.05
gaseous absorption : O3 (TOMS), NO2 (climato), H20 (meteo)
accuracy : typically 2% (3% in blue)
Rayleigh
ocean
aerosols
molecular
aerosol
marine
gaseous
I_mean
443
84.25
1.25
14.48
-0.56
0.1177
490
85.25
1.98
12.75
-1.84
0.0842
565
90.56
3.76
5.67 -8
670
90.23
7.5
2.25
-3.67
Main contributors
to TOA reflectance
(in %)

5. Method Analysis over predefined and characterized oceanic sites
selection of candidate sites
through a climatology based on SeaWiFS data
spatial homogeneity for each site
limited (=“controlled”) seasonal variation
Benefit to calibrate over various oceanic sites
1 site = still a small possible bias due to exact knowledge of rw
statistical approach : distinguish sensor from sites behaviors
analysis of correlations with various geometrical or geophysical parameters
different for each sites (Latitude / monthly variations)
strong credible if a phenomenon is observed for each site (ex: time evolution)
more accurate analysis
443 – Arbitrary value
443 - Climatology

6. Dedicated sites Step 1 = geographical selection ClimZOO :
Climatology of Oligotrophic
Oceanic Zones
Measurement Selection
(POLDER3 - 14,000 points
from Jan 05 to Sep 06)
(from Fougnie et al., 2002 and 2010)

7. Selection stream
Step 2 : some sites may be not recommended for some months
refer to the climatologic table
Step 3 = rough cloud mask
typically using level-1 mask or simple threshold : quite easy over ocean
Step 4 = cloud mask dilatation
reject all pixels in the vicinity of clouds (ideally at least 30km)
Step 5 = geometrical selection
avoid data on the sunglint region : predictable using the wave angle (<30°)
Step 6 = surface wind speed
avoid sea surface foam : discard wind speed over 5 m/s (ideally)
Step 7 = quality index control
discard all pixels corresponding to inappropriate quality index (depending on the sensor)
Step 8 = invalid ancillary data
discard data corresponding to no or abnormal WV, Ozone, Pressure, wind speed
Step 9 = “turbid” pixel rejection
threshold on NIR band : radiance(NIR)*cos(VZA)*cos(SZA) < (ideally)
discard : aerosol load, residual clouds, residual surface foam, UFO

8. Computation of the TOA signal
for the geometrical conditions of the considered pixel (elementary measure) :
gaseous
transmission
atmospheric
scattering
marine
reflectance
atmospheric
diffuse
transmission
aerosol
scattering + coupling
Rayleigh
scattering

9. Computation of the TOA signal
Inputs :
IRay, Iaer, Ttot are computed using an accurate radiative transfer code
Successive Order of Scattering (SOS) : coupling, polarization
require the knowledge of the Rayleigh equivalent optical thickness for the considered spectral band
Iaer is evaluated using the NIR band after Rayleigh correction (black surface)
because turbid cases were rejected, only very small loads are observed = residual background
a Maritim-M98% mode is considered to extrapolate to shorter bands
if a calibration error is know for the NIR band – an iteration is necessary
Gaseous transmission are evaluated using SMAC (simplified model from 6S)
Rw is taken from a climatology – interpolation for the spectral bands

10. Unified computation This is an absolute calibration method
means, we don’t use reference band or reference sensor
at the first order, the reference is the Rayleigh scattering
Rayleigh scattering is not perfectly know :
1% uncertainty on the computation of the molecular optical thickness
uncertainty due to the radiative transfer code : multiple scattering, polarization
Compare different sensor using Rayleigh scattering requires to use :
at least exactly the same “Rayleigh” reference
ideally all the same contributors
The current reference method is Rayleigh Calibration Version 3.5
See development plan

11. Development Plan
The reference method is Rayleigh Calibration Version 3.5
See development plan (GSICS inspired graphical representation)
DEV = study & ATBD first definition [resp. SI/MO]
PROTO = prototype on dedicated test environment on MUSCLE – Final ATBD [resp. SI/MO]
Pre-OPE = test on the operational MUSCLE [resp. ME/EI]
OPE = fully operational method / Traceability guaranteed [resp. ME/EI]

12. A quick look on the Error Budget
The elementary calibration error :
the radiative transfer code inaccuracy
Rayleigh scattering knowledge
inputs / ancillary data uncertainty
aerosol contribution (use of NIR info)
marine reflectance
Statistical approach – Matchups
 Robust error budget // uncertainty analysis is under construction
Results are very sensitive to :
knowledge of the spectral response and its variation inside the field-of-view
polarization sensitivity of the instrument
possible non-linearity for very dark signal, stray light…
 Consequently : it is a very good diagnostic

13. A quick look on the Error Budget
Contributors to the
computed calibration
SADE

14. A quick look on the Error Budget
Computation of error factors
Calibration coefficient :
Derivate :
Error factor

15. A quick look on the Error Budget
Level-1
Error sources tree
Error factors
Level-2

16. A quick look on the Error Budget
Error budget construction
document the variability range for each source :
strongly depend of the considered sensor
necessary to compute the error factor for each source
document the uncertainty for each source
necessary to compute the error for each source
On-going
Produce the error budget for
MERIS : OC profile
PARASOL : bidirectional sensor
Végétation : broadband not systematic acquisitions
SEVIRI : GEO representative

17. Examples of Application
The Rayleigh scattering method was deployed :
historical development with narrow band sensors : POLDER 1&2
and broad band sensors : Végétation 1 & 2
dedicated ocean color sensor : SeaWiFS (US), MERIS (ESA)
high-resolution imagery sensor : SPOT6, Pleiades
preliminary over Geostationary sensor : SEVIRI
planned : MODIS, Sentinel-3 (OLCI/SLSTR), Sentinel-2

18. Why not immediately applicable to GEO
The principle will be exactly the same for GEO : selection + calibration
But this experience cannot be immediately generalized to GEO
Statistically :
only a few sites are accessible : equilibrium North/South + various oceans
range of geometries are not the same : ~fix VZA, mainly variations with solar geomerty
influence of NIR band (not designed for very low signal)
Need to validate on a large dataset
but very promising results for SEVIRI : some signature to understand (backscattering)
GEO
LEO

19. Why not immediately applicable to GEO
range of geometries are not the same : ~fix VZA, mainly variations with solar geometry
LEO - SeaWiFS
GEO - SEVIRI
Small VZA population
Backscattering / principal plane population

20. References Methodology : Application :
Fougnie et al., 2002, Identification and Characterization of Stable Homogeneous Oceanic Zones : Climatology and Impact on In-flight Calibration of Space Sensor over Rayleigh Scattering, Ocean Optics XVI Proceedings
Fougnie et al., 2010, Climatology of Oceanic Zones Suitable for In-flight Calibration of Space Sensors, Earth Observing Systems XV, SPIE Optics & Photonics
Llido et al., 2010, Climatology of Oceanic Zones Suitable for In-flight Calibration of Space Sensors, Rapport d’étude CNES
Application :
Fougnie et al., 2007, PARASOL In-flight Calibration and Performance, Applied Optics
Hagolle et al., 2006, Meris User Meeting
Jolivet et al., 2009, In-flight Calibration of Seviri Solar Channels on board MSG Platforms, Eumetsat User Meeting
Fougnie et al., 2012, “Validation of the MERIS 3rd Reprocessing Level-1 Calibration Through Various Vicarious Approaches – Perspectives for OLCI,“ Sentinel-3 OLCI/SLSTR and MERIS/AATSR Workshop
Fougnie B. et al., 2012, “In-Flight Calibration of Space Sensors Through Common Statistical Vicarious Methods : Toward an Ocean Color Virtual Constellation,” Ocean Optics XXI