Moon Lunar irradiance must be Stable

Moon Lunar irradiance must be: • Stable – Assumed • Known – ROLO • Measured – Instrument • Available – How frequently X. Wu, T. C. Stone, F. Yu, and D. Han, "Vicarious calibration of GOES Imager visible channel using the Moon“. Proc. SPIE, Vol. 6296, dio: 10. 1117/12. 681591. 9 -11 Feb 2010 Toulouse, France GRWG-5 1

The Moon as an on-orbit radiometric reference The Moon is a “solar diffuser” with exceptional stability properties, <10− 8 per yr v the inherent stability of the Moon enables robust modeling § a model is required to predict the lunar brightness for any observation of the Moon from a spacecraft platform § thus the model is the lunar reference The USGS ROLO program has built a successful technique for using the Moon v developed under NASA research sponsorship for EOS instruments (MODIS, Sea. Wi. FS, ALI, Hyperion, ASTER, MISR, . . . ) v utilized by NOAA for geostationary visible-channel imagers § regular dedicated Moon observations by GOES−E and −W § post-launch test observations by GOES− 13 and − 14 v currently operates using interactive data exchanges with USGS (T. Stone) A lunar component is included in the calibration plans for upcoming operational satellites (LDCM, GOES−R, METEOSAT 3 rd Gen. ) 9 -11 Feb 2010 Toulouse, France GRWG-5 2

Capabilities of Lunar Calibration v Sensor response stability § using a time-series of Moon observations § capable of achieving <0. 1% per year stability, demonstrated by Sea. Wi. FS § this meets stability requirements for climate change measurement from space v Back-calibration of images of the Moon § the lunar model can be applied to Moon observations made at any time § currently being applied to the ISCCP GEO archive § future improvements to the lunar reference can be applied to current & past data v Inter-calibration of sensors that view the Moon § the spectral consistency of the Moon allows cross-calibration of sensors with overlapping but slightly different bands § cross-calibration of Sea. Wi. FS and MODIS has been investigated § cross-calibration of the ISCCP GEO visible imagers currently in progress v Future capability: the Moon as an absolute reference § efforts by USGS and NIST to achieve SI traceability w/ absolute uncertainty <1% 9 -11 Feb 2010 Toulouse, France GRWG-5 3

Steps of Lunar Calibration v Observe the Moon v Know what should be measured – EModel(t) v Know what have been measured – EGOES(t) v Fit EGOES(t)/ EModel(t) = Aeβt 9 -11 Feb 2010 Toulouse, France GRWG-5 4

Many complete images of gibbous moon v Easier for GEO § No spacecraft maneuvers v No longer a problem for scheduled collection v Remain a problem even for scheduled collection § A few per year unscheduled § A dozen per year scheduled § May need ~100 per year – likely for GOES-R 9 -11 Feb 2010 Toulouse, France GRWG-5 5

Steps of Lunar Calibration v Observe the Moon v Know what should be measured – EModel(t) v Know what have been measured – EGOES(t) v Fit EGOES(t)/ EModel(t) = Aeβt 9 -11 Feb 2010 Toulouse, France GRWG-5 6

ROLO, a sophisticated photometric model of lunar reflectance, precision approaching 1% Ak – disk equivalent reflectance 9 -11 Feb 2010 Toulouse, France GRWG-5 7

Steps of Lunar Calibration v Observe the Moon v Know what should be measured – EModel(t) v Know what have been measured – EGOES(t) v Fit EGOES(t)/ EModel(t) = Aeβt 9 -11 Feb 2010 Toulouse, France GRWG-5 8

Computation of Lunar Irradiance i: Index of Moon pixel Ri: Radiance from pixel i ωi: Solid angle subtended by pixel i S: Prelaunch cal coefficient (slope – reciprocal of instrument gain) Ci. R: Raw count of pixel I CS: Space count relevant to Ci. R EGOES = ωS∑i(Ci. R – CS) 9 -11 Feb 2010 Toulouse, France GRWG-5 9

Which are lunar pixels Take all, since space pixels contribute nothing 9 -11 Feb 2010 Toulouse, France GRWG-5 10

Variable amount of stray light Strong Asymmetric 9 -11 Feb 2010 Toulouse, France Weak Absent GRWG-5 11

Ellipse fitted to smoothed moon 9 -11 Feb 2010 Toulouse, France GRWG-5 12

Computation of Lunar Irradiance i: Index of Moon pixel Ri: Radiance from pixel i ωi: Solid angle subtended by pixel i S: Prelaunch cal coefficient (slope – reciprocal of instrument gain) Ci. R: Raw count of pixel I CS: Space count relevant to Ci. R EGOES = ωS∑i(Ci. R – CS) 9 -11 Feb 2010 Toulouse, France GRWG-5 13

Moon is dark – mean ρ ~5% ~1 count drift can be up to 2% signal 9 -11 Feb 2010 Toulouse, France GRWG-5 14

Single detector line mean space count 9 -11 Feb 2010 Toulouse, France GRWG-5 15

Computation of Space Count 9 -11 Feb 2010 Toulouse, France GRWG-5 16

Steps of Lunar Calibration v Observe the Moon v Know what should be measured – EModel(t) v Know what have been measured – EGOES(t) v Fit EGOES(t)/ EModel(t) = Aeβt 9 -11 Feb 2010 Toulouse, France GRWG-5 17

All pixels, selected mean 9 -11 Feb 2010 Toulouse, France GRWG-5 18

Deep Convective Clouds (DCC) Assumption 9 -11 Feb 2010 Toulouse, France GRWG-5 19

Detection of DCC Pixels Criterion for selecting DCC: • Tb(10. 7) < 205 o. K • STDV [Tb(10. 7)] < 1 o. K • STDV [R(0. 65)] < 2% • 20 N-20 S, 45 W-120 W Area for STDV [Tb(10. 7)] Viewing Geometry constraints: SZA< 60 VZA < 40 RAA: 10 ~170 Area for STDV [R(0. 65)] 9 -11 Feb 2010 Toulouse, France GRWG-5 20

GOES 10 Data v GOES 10 (W) Imager visible channel lifetime data: 07/07/1998 to 06/27/2006 are used to search for DCC pixels within geographic area constraints v For each day image taken around local noon (~21 UTC) is examined for DCC pixels v Mean reflectances are calculated for each month v Mean Reflectances are normalized to reduce effect of seasonal fluctuations (Chung, CALCON 2009) 9 -11 Feb 2010 Toulouse, France GRWG-5 21

Normalization of Reflectance Rnorm = [(Cscene-Cspace)*m/cos(SZA)]*DSE 2/ADM Rnorm : Normalized reflectance C : Digital count m : pre-launch calibration coefficient SZA : solar zenith angle DSE : Sun-Earth distance (unit: AU) ADM : Angular Distribution Model - CERES/TRMM ADM (NASA) - Bidirectional Reflectance Distribution Function (BRDF) (B. J. Sohn) 9 -11 Feb 2010 Toulouse, France GRWG-5 22

Angular Distribution Model (ADM) Solar Zenith Angle = 30 -40 – - Shortwave ADM for ice cloud over ocean – - cloud optical depth > 50 BRDF • - computed by SBDART radiative transfer model • - use cloudy scattering properties data • Baum et al. (2005) phase function for nonspherical ice particles • Mie phase function for water particles 9 -11 Feb 2010 Toulouse, France CERES ADM Anisotropic Factor CERES/TRMM ADM BRDF phi = 90 -180 phi = 0 -90 Satellite Zenith Angle ( ) BRDF CERES/TRMM ADM GRWG-5 23

Degradation of Mean Reflectance without normalization 95% confidence interval Correlation coefficient =0. 909 SSE=3. 4% Slope=0. 0495+/-0. 0024 Coeff=78. 01+/-0. 77 Mean. Refl = coeff*e-slope*t 9 -11 Feb 2010 Toulouse, France GRWG-5 24

Degradation of Mean Reflectance normalized with CERES/TRIMM ADM Correlation coefficient =0. 952 SSE=2. 8% Slope=0. 0495+/-0. 0017 Coeff=93. 02+/-0. 64 Mean. Refl = coeff*e-slope*t 95% confidence interval 9 -11 Feb 2010 Toulouse, France GRWG-5 25

Degradation of Mean Reflectance normalized with BRDF 95% confidence interval Correlation coefficient =0. 973 SSE=2. 04% Slope=0. 0503+/-0. 0013 Coeff=89. 28+/-0. 46 Mean. Refl = coeff*e-slope*t 9 -11 Feb 2010 Toulouse, France GRWG-5 26

9 -11 Feb 2010 Toulouse, France GRWG-5 27

9 -11 Feb 2010 Toulouse, France GRWG-5 28

Comparison and Discussion Independent Stable Available Properly Observed Traceable Latency Cost EDF 1 1 5 5 1 1 4 Desert 1 2 5 5 2 1 5 Star 5 5 3 2 1 1 3 MODIS 5 4 3 5 5 5 1 Scores are comparative. 1=Worst, 5=Best Green: Quality considerations. Orange: Operation considerations 9 -11 Feb 2010 Toulouse, France GRWG-5 29

Summary (star) v Navigation data can be used for calibration § Substantial reprocessing is necessary § A comprehensive plan benefits both navigation and calibration (started 2008) v Intra-annual variation is the limiting factor § Theoretical understanding § Empirical algorithm v A new algorithm seems promising v Recommendations made to GOES-R System Engineer through Calibration Working Group 9 -11 Feb 2010 Toulouse, France GRWG-5 30

MODIS v Pros § Absolute calibration for GOES § Frequency in GOES-MODIS collocation events(~3 -4 times/month) § Calibration uncertainty < 5% v Cons § EDF match method is sensitive to distribution of cloud reflectance frequency histogram • Difference in SRFs may cause relatively large EDF difference for the low cloud coverage pixels. § After almost +7. 5 years in mission, MODIS could be out of service soon 9 -11 Feb 2010 Toulouse, France GRWG-5 31

Summary (moon) v Lunar calibration for current GOES is feasible but noisy v Stray light seems not as critical as previously thought v Knowledge of space count seems the limiting factor v More promising for GOES-R § Enhanced bit-depth § More observations 9 -11 Feb 2010 Toulouse, France GRWG-5 32

Conclusions (DCC) DCC are an excellent target for trending of GOES Imager visible channel: v Enough of DCC pixels are found from one image per day to produce meaningful and stable statistics v DCC mean monthly reflectance normalized with BRDF works best for trending of visible channel degradation v Yielded degradation rate of 5. 0% with 2% standard error of estimate and 2. 8% precision v Similar to Lunar calibration results with somewhat better accuracy 9 -11 Feb 2010 Toulouse, France GRWG-5 33

Discussion – EDF v The independence of the reference is highly questionable (better if monitoring surface features) v The stability of the reference is highly questionable v Required data are always available v Date collection is identical to those for other earth view data v The absolute value of the reference is unknown v Takes years for calibration to start v Development/operation/maintenance cost: med/low 9 -11 Feb 2010 Toulouse, France GRWG-5 34

Discussion – Desert v The independence of the reference is highly questionable (better if monitoring atmospheric features) v The stability of the reference is questionable v Required data are always available v Date collection is identical to those for other earth view data v The absolute value of the reference is not well know v Takes years for calibration to start v Development/operation/maintenance cost: med/low 9 -11 Feb 2010 Toulouse, France GRWG-5 35

Discussion – Star v The calibration reference is totally independent of the object being observed v The reference is extremely stable v Annual data gap (a few months each year) v Different date collection § Additional electronic circuitry § Light source does not cover the full aperture § Variety of color temperature (can be an advantage) v The absolute value of the reference is unknown v Takes years for calibration to start v Labor-intensive in development but can be automated in operation v Development/operation/maintenance cost: hi/med 9 -11 Feb 2010 Toulouse, France GRWG-5 36

Discussion – MODIS v The calibration reference is totally independent of the object being observed v The reference is expected to be very stable v Required data are available ~once a week. In addition, MODIS is approaching the end of its designed life v Date collection is identical to those for other earth view data v The absolute value of the reference is well know v Calibration can start immediately after launch v Development/operation/maintenance cost: hi/hi/med 9 -11 Feb 2010 Toulouse, France GRWG-5 37

Recommendation v All methods be encouraged § Can afford § Likely beneficial v Concerted effort to comprehensively characterize DCC with MODIS § § Spatial Temporal Spectral Angular 9 -11 Feb 2010 Toulouse, France GRWG-5 38

Summary v GOES Imager visible channel needs regular vicarious calibration. v Four options have been explored: § § Empirical Distribution Function (EDF) Desert Star Observations MODIS v The four options were compared in terms of quality and operation v MODIS-based calibration is ready for operation v Star-based monitoring may have other values v Recommend to implement the MODIS-based calibration while keeping the other options for validation and other purposes 9 -11 Feb 2010 Toulouse, France GRWG-5 39