1. On the use of Ray-Matching to transfer calibration
D. Doelling, P. Minnis
NASA LaRC
Raj Bhatt, Dan Morstad
SSAI
GSICS/GWRG Session Daejon, Korea, March 29-April 1, 2011

2. GEO to MODIS Cross-Calibration Method
Ray-match coincident GEO counts (proportional to radiance) and MODIS radiances
use a 0.5°x0.5° lat by lon grid to mitigate navigation and time matching errors
Use MODIS as reference since GEOs have no onboard calibration
Normalize solar constants and SZA, obtain MODIS equivalent radiance
Perform monthly GEO/MODIS regressions of the gridded radiances, and derive monthly gains
Compute timeline trends from the monthly gains

3. GOES-12/Terra-MODIS GOES-12 gain GOES-12/Terra-MODIS
July 2003
GOES-12 gain
based on Terra-MODIS
Gain = 0.68
29 published count offset

4. Remove SRF difference between GEO and MODIS using pseudo SCIAMACHY radiances
SCIAMACHY is a imaging spectrometer designed to measure trace gas absorption
3200 wavelengths between 0.25 and 1.75µm in 6 bands, which are normalized to each other
Use nadir footprints, size= 32 (along track) x215km
ENVISAT, local time, launched March 2002
Compute SRF adjustment factors
Regression from all SCIAMACHY footprint pseudo radiances over the GEO/MODIS calibration domain
Use spectral response function from GEO and MODIS
Validate
Cross calibrate GEO and MODIS over land and ocean
Derive SCIAMACHY adjustment factors for land and ocean separately
The gains for ocean and land should be similar

5. GOES-12/Terra-MODIS, Sept. 2009 no spectral correction
Ocean
Land
Gain =
Offset = 31.4
Gain =
Offset = 47.0
• GEO gain dependent on instrument spectral response and scene type
• Note surface type effects mainly the offset under clear-sky conditions
• The gain difference is 5.3%, and the offset should be 29.0

6. SCIAMACHY* spectra Ocean ForestG
Terra 24.6 G
Terra 17.4
• Clear-sky SCIAMACHY mean & sigma spectral response over ocean and forest
• Compute the pseudo G12 and MODIS radiance from SCIAMACHY
*Courtesy of SCIAMACHY team

7. SCIAMACHY* spectra Low Cloud High Cloud
Terra
Low Cloud
High Cloud G
Terra
• Bright cold high clouds have radiance ratios near one
• Bright low clouds have more absorption in the near IR

8. SCIAMACHY spectral corrections, July 2003
Ocean (All months)
Land (Oct)
• Use all SCIAMACHY footprints that fall into the GEO equatorial domain during
• Compute pseudo GOES and MODIS radiances using SCIA spectra
• Derive spectral correction using a cubic fit for ocean and water

9. GOES-12/Terra-MODIS, July 2003 with spectral correction
Ocean
Land
Gain =
Offset = 30.7
Gain =
Offset = 29.8
• The gain difference before spectral correction = 5.3%, offset=31.4, 47.0
• With spectral correction the gain difference = 0.3%, offset close to 29

10. Which MODIS for reference?
Use MODIS as an absolute reference for calibration transfer using ray-matching
Use desert and DCC to determine, which MODIS is more stable
Libya, Egypt, Arabian, Dome C, Greenland, Sonora used
Also Terra/Aqua NSNOs indicates that there is relative trending between the two
MODIS instrument team indicates that Aqua is better characterized
MODIS absolute calibration error is ~2%

11. 10-year Clear-sky CERES SW BB monthly sigma
Terra-CERES (%) (10:30AM)
Aqua-CERES (%) (1:30PM)
• ISCCP uses the entire clear-sky earth as a calibration target for AVHRR
• Use the CERES clear-sky SW (BB) deseasonalized monthly radiance to derive a seasonal climatology, then take the standard deviation of each of 120 months subtracted from climatology.
• Sahara, especially Egypt has the lowest variability, Arabia, Greenland and Antarctica are stable targets

12. LaRC MODIS desert methodology, Arabia
10%-
5%-
0%-
BRDF sigma, vz=25°
RAZ bins
SZA bins
Daily spatial sigma radiance
Daily TOA radiance
• Use a spatial sigma threshold to identify clear-sky, not necessarily pristine
• Deseasonalize monthly mean radiances, by using 12 month running means, or derive trends for each month and average trends for overall trend
• Or use BRDF using a 3 year stable time period

13. LaRC MODIS desert methodology, Arabia
BRDF Monthly
Normalized trend
Deseasonalized
Normalized trend
BRDF Daily
TOA Radiance
Stderr=1.3%
Gain=0.2%/decade
Stderr=1.0%
Gain=0.15%/decade
• Both methods reveal that Arabian desert site is observed with Aqua is stable
• The methods are very consistent

14. Desert calibration using site specific TOA BDRF
Terra-CERES 0.65µm
Aqua-CERES 0.65µm
Range -1.1 to -1.9,
-1.5 mean %/decade
Range -0.9 to +0.4,
-0.3 mean %/decade
Terra DCC
Aqua DCC
• Desert and DCC indicate Aqua more stable than Terra
• Use Aqua as reference

15. Radiometrically scale Terra to Aqua MODIS
Remove calibration differences between Terra and Aqua-MODIS, use Aqua-MODIS as reference
Use Terra and Aqua NSNO (within 15 minutes) to cross-calibrate Terra and Aqua
Assume Terra and Aqua have same SRF
Correct Terra radiances to match Aqua
Use simple gain factor
Both GEO/Terra and GEO/Aqua ray-matching should provide independent reference
If ray-matching is robust, both Terra and Aqua based calibration transfer should be identical

16. Terra/Aqua MODIS 0.65 µm NP, July 1, 2004
Similar for Terra/Aqua at 71°N
0.65 µm NP, July 1, 2004
• nadir
• off-nadir
• Increase dynamic range by using off-nadir 50km regions
• Use 50km regions to mitigate time mismatch
• Off nadir regions along the longitude of constant VZA are at solar noon and have similar RAZ
• Off nadir slopes are nearly identical to nadir slopes

17. Terra/Aqua Collection 5
Band1 (0.65µm)
Band1 (0.65µm)
Nov 2003
Apr 2009
Terra to Aqua adjustments applied to collection 5
1.011
1.019
1.032
• Terra/Aqua relative calibration, shows two major discontinuities
• Use 3 adjustment factors
• Off-nadir is less noisy over time

18. Terra-MODIS Band1 Collection 5 to 6 adjustment factors
Mirror side 0
Mirror side 1

19. Terra/Aqua Collection 6
Band1 (0.65µm)
Band1 (0.65µm) corrected
adjustments applied
For collection 6
1.026
7.804e-6*DSL+1.008
2.8%/decade

20. Compare Aqua-MODIS/GEO, Terra-MODIS/GEO, desert and DCC calibrations
First, determine if stability is consistent between all methods
Second, determine the absolute gain difference
SRF correction between MODIS and historical GEO sensors can be on order of 5%
Ray-matching relies on bright cloud targets
Offset is from space count
Bright clouds are spectrally flat at surface
However atmospheric correction can vary with cloud height and season, DCC~1%, but stratus ~5%
Need to develop scene dependent SRF adjustments

21. GOES-12/MODIS Collection 5
Before Terra Adjustment
After Terra Adjustments
• DCC and deserts normalized to Terra
• Note GEO/Terra diverging after 2009 from DCC and desert before adjustment

22. Met-9/MODIS collection 5
Before Terra Adjustment
After Terra Adjustment

23. MET-9/MODIS Collection 5 relative
Before TA adjustment
After TA adjustment
• Normalized gains

24. MET-7 Libya desert Daily TOA radiance Daily BRDF TOA radiance
Monthly BRDF TOA radiance
Monthly BRDF TOA radiance after SCIAMACHY correction
10% variation due to atmosphere

25. Comparison of MET-7 with MODIS, DCC and Desert
• SCIAMACHY correction for Desert and DCC
• Absolute calibration of DCC and desert based on Aqua
• GEO/MODIS not spectrally corrected

26. GEO/MODIS variability
MET-7/MODIS 0.65µm pseudo radiances from SCIAMACHY
Met-8/MODIS0.86µm pseudo radiances from SCIAMACHY
Ice
water
~5%
~10%
• Seasonal cycle due to looking at various scenes
• Assume bright cloud surface is spectrally flat
• Need to take out atmosphere, depending on scene type
• Use scene dependent SCIAMACHY models

27. Conclusions Ray-matching calibration transfer accuracy depends
Stability of reference sensor (systematic)
SRF correction (systematic)
Matching errors due navigation, time difference, and angular thresholds (random noise)
Ray-matching improvements
Use Aqua Collection 6 as reference
Radiometrically scale Terra to Aqua, two independent references
Use of SCIAMACHY for SRF correction
Ray-matching needed improvements
Scene dependent SRF corrections
Verify using other methods, such as DCC and deserts