1. TOA Radiative Flux Estimation From CERES Angular Distribution Models
Norman G. Loeb
Hampton University/NASA Langley Research Center
Hampton, VA
Acknowledgements: K. Loukachine, S. Kato, N. M. Smith
August 28, 2002

2. Outline Introduction TOA flux retrieval strategy – ADM definition
ADM examples for various Earth scenes
observed by CERES on TRMM and Terra
4. TOA flux validation results
5. Summary

3. “Heat balance of the Atmosphere” (Dines, 1917)

4. Global Radiation Budget

6. CERES Instrument
5 instruments on 3 satellites (TRMM, Terra, Aqua) for diurnal and angular sampling.
Narrow field-of-view scanning radiometer with nadir footprint size of 10 km (TRMM); 20 km (Terra & Aqua).
Measures radiances in mm, mm and 8-12 mm.
Capable of scanning in several azimuth plane scan modes: fixed (FAP) or crosstrack, rotating azimuth plane (RAP), programmable (PAP).
Coincident Cloud and Aerosol Properties from MODIS/VIRS

7. CERES Azimuth Plane Scan Modes
Fixed (FAPS)
CERES Azimuth Plane Scan Modes
Programmable (PAPS)
Rotating (RAPS)

9. Instantaneous Fluxes at TOA and Angular Distribution Models
CERES Radiance Measurement
TOA Flux Estimate SW LW WN

10. Instantaneous Fluxes at TOA and Angular Distribution Models
TOA flux estimate from CERES radiance:
where,
Rj (o , ,) is the Angular Distribution Model (ADM) for the “jth” scene type.

12. ADM Scene Identification
The main reason for defining ADMs by scene type is to reduce the error in the albedo estimate.
=> Earth scenes have distinct anisotropic characteristics which depend on their physical and optical properties. (e.g. thin vs thick clouds; cloud-free, broken, overcast etc.).
=> Scene identification must be self-consistent. Biases in cloud property retrievals (e.g. due to 3D cloud effects) should not introduce biases in flux/albedo estimates.

14. ADM Scene Identification
The main reason for defining ADMs by scene type is to reduce the error in the albedo estimate.
=> Earth scenes have distinct anisotropic characteristics which depend on their physical and optical properties. (e.g. thin vs thick clouds; cloud-free, broken, overcast etc.).
=> Scene identification must be self-consistent. Biases in cloud property retrievals (e.g. due to 3D cloud effects) should not introduce biases in flux/albedo estimates.

15. Anisotropic Model Scene Type Stratification

17. Scene Types for CERES/TRMM SW ADMs

18. Scene Types for CERES/TRMM LW and WN ADMs
ADM Category
Parameter Stratification
Total
Clear
Ocean
3 Precipitable Water 15
5 Vertical Temperature Change
Land
Desert
Broken Cloud Field
(4 intervals)
Ocean/Land/Desert
288 (O)
288 (L)
288 (D)
6 DT (Sfc-Cloud)
4 IR Emissivity
Overcast
Ocean+ Land+Desert
126
7 DT (Sfc-Cloud)
6 IR Emissivity

19. Observations 1) CERES/TRMM: 9 months of CERES/TRMM SSFs:
January-August March 2000
9 months of CERES/TRMM ES8s: CERES “ERBE-Like” product.
2) CERES/Terra:
Four months of CERES/Terra SSFs
Nov-Dec Apr-May 2001
- Anticipate constructing CERES/Terra ADMs with 2 years of SSFs!

28. New ADMs for Terra: Approaches Being Considered
1. Shortwave:
Increase resolution of angular bins to 2
Clear Ocean: Similar approach as on CERES/TRMM
=> Wind speed-dependent ADMs with theoretical correction for aerosol optical depth variations.
Clouds over Ocean: Continuous scene type using sigmoidal functional fits to data.
Clear Land: Stratify by IGBP type + vegetation index + taer
=> Is there any change in anisotropy?
Clouds over Land: Continuous scene type using sigmoidal functional fits to data.
Clear Snow: Stratify by NDSI
Clouds over Snow:
2. Longwave and Window:
Similar to CERES/TRMM but at higher angular resolution
Empirical ADMs over snow

37. Klein et al., 2002

38. NDSI
NDSI

40. SW TOA Flux Validation
Does mean all-sky flux depend on viewing geometry?
Comparisons with Direct Integration Fluxes:
Solar zenith angle dependence (SW)
Latitudinal dependence
Regional fluxes
Instantaneous Flux Uncertainties
Use alongtrack data to examine the self-consistency of instantaneous TOA fluxes from a scene for different viewing geometries.

46. Mean LW Flux vs Viewing Zenith Angle (Jan-Mar 1998)
Daytime
Nighttime

47. Flux Bias Determination
ADM mean flux bias in angular bin (qo ,qj ,fk) :
Footprint-weighted ADM mean flux bias:
where wj is a weighting factor accounting for the relative effect of different viewing zenith angles on gridded time-averaged fluxes.
nk and nj are the number of relative azimuth and viewing zenith angle bins.

49. Solar Zenith Angle Distribution by Latitude (March 1998)
Relative Frequency (%)

50. ADM Mean Regional Flux Biases (q < 70°)
(March 1998 Solar Zenith Angle Sampling)
CERES ERBE-Like – DI Flux Difference
CERES SSF – DI Flux Difference
Flux Difference (W m-2)

52. Daytime LW ADM Mean Regional Flux Biases (q < 70°)
(Jan, Feb, Mar 1998)
ERBE-Like – DI Flux Difference
SSF Ed2 – DI Flux Difference
Flux Difference (W m-2)

54. (Ocean Ocean; qo=42° - 44°; q < 25°)
ERBE-Like – CERES SSF SW Flux Difference vs Cloud Optical Depth
(Ocean Ocean; qo=42° - 44°; q < 25°)
Liquid Water Clouds
Ice Clouds
SW Flux Difference (W m-2)
Cloud Optical Depth
Cloud Optical Depth
Number of FOVs
Number of FOVs

55. 1 Regional Instantaneous SW TOA Flux Consistency Test
Objective: Compare ADM-derived TOA fluxes over 1 regions from different viewing geometries. Are TOA fluxes consistent?

56. 1 Regional Instantaneous SW TOA Flux Consistency Test
Approach: For every 1 region:
Use nadir CERES footprints to train MODIS imager to produce broadband SW TOA fluxes over each CERES footprint within 1 region (linear fit).

57. (2) Convert mean imager visible radiance over every CERES alongtrack footprint to a broadband flux using fit in (1).
=> Imager sees CERES alongtrack footprints from nadir direction only.
(3) For every 1 region, compare nadir imager TOA fluxes with alongtrack CERES TOA fluxes.
=> Since imager and CERES measurements are collocated, spatial matching errors are reduced with this technique.
=> Main Sources of Error:
i) Radiance-to-flux conversion (ADMs)
ii) Narrow-to-broadband conversion

58. 1 Regional Instantaneous SW TOA Flux Consistency Test
Results: Based on 1 day of A-track and X-track CERES/Terra data.

60. Summary CERES goes well beyond ERBE:
i) Coincident imager-based cloud and aerosol properties together with broadband CERES radiative fluxes.
ii) Improved accuracy of TOA fluxes by a factor of 2-4.
Challenges:
Improve CERES/Terra and CERES/Aqua TOA flux accuracy over CERES/TRMM.
Validation: - Need quantitative estimates of errors.
- Comparisons with other instruments (MISR, POLDER, GERB).
Climate model use of global datasets, comparisons between models and observations; improved model parameterizations.

61. Future:
Merging of measurements from CERES & MODIS (Aqua)
with CALIPSO, CloudSat, PARASOL.

62. Backup Slides

63. ADM Scene Surface Types
ADM Scene Type

64. SW ADM Frequency of Occurrence by
Cloud Fraction & Adjusted Cloud Optical Depth
(Ocean) h
Liquid Water Clouds
Ice Clouds
Adjusted Cloud Optical Depth
Cloud Fraction (%)
Cloud Fraction (%)