Towards Improved CERES Angular Distribution Models Wenying Su Lusheng Liang Joseph Corbett Zachary Eitzen SSAI, Hampton VA Thanks to Norman Loeb! 4/26/11
 CERES
STM
 0 Outline •  Introduction and how to evaluate the angular distribution model (ADM) error; •  Shortwave clear-sky ADM over land/desert, fresh snow; •  Shortwave clear-sky ADM over ocean; •  Shortwave clear-sky ADM over permanent snow and sea ice; •  Longwave daytime cloudy-sky ADM over ocean. 4/26/11
 CERES
STM
 1 Radiance and flux •  CERES measures radiance: I(θ , θ, ϕ) 0 •  Relationship between radiance and flux: θ θ 0 ϕ 4/26/11
 CERES
STM
 2 The road from radiance to flux: angular distribution mode l •  Sort observed radiances into angular bins over different scene types; •  Integrate over all θ and ϕ to estimate the anisotropic factor for each scene type; •  Apply anisotropic factor to observed radiance to derive TOA flux; 4/26/11
 CERES
STM
 3 Predicted radiance vs. observed radiance •  Predicted radiances can be used to verify the accuracy of ADM; 4/26/11
 CERES
STM
 4 Normalize predicted and observed radiance Observed radiance: 1° Predicted radiance: 1° •  RMS error between normalized predicted radiance and normalized observed radiance is closely related to the ADM error 4/26/11
 CERES
STM
 5 Clear-sky land/desert angular distribution model •  Collect clear-sky CERES reflectance over 1°✕1° regions for each month; •  Stratify reflectance within each 1°✕1° region by NDVI (0.1) and cosθ (0.2); 0 •  Apply an 8-parameter fit to produce BRDF and ADM for each NDVI and cosθ intervals within each 1°✕1° region. 0 From Ahmad & Deering, 1992 4/26/11
 CERES
STM
 6 Use RMS error between normalized predicted and measured radiance to test the ADM RMS error (%) using Ed2 ADM for 200305 FM2 over clear-sky land/desert: Mean RMS error = 6.1% % 4/26/11
 CERES
STM
 7 A simpler BRDF model: Modified RossL i •  B1 estimates the directional reflectance of a flat surface with randomly distributed and oriented protrusions; •  B2 approximates the radiative transfer within a vegetation canopy, accounts for the hot spot effect; from Maignan et al., 2004 Kernel B1 Kernel B2 4/26/11
 CERES
STM
 8 The new BRDF model reduces the RMS error RMS error (%) using prototype Ed4 ADM for 200305 FM2 over clear-sky land/desert: Mean RMS error = 5.5% % 4/26/11
 CERES
STM
 9