An improved derivation of the top-of-atmosphere albedo from POLDER/ADEOS-2: 2. Broadband albedo

The narrowband albedos derived from Polarization and Directionality of Earth Reflectances (POLDER) measurements have been described in a companion paper (Buriez et al., 2005). Here, they are used to estimate the broadband shortwave albedo. Except for the gaseous absorption, the albedos at 443 and 670 nm are considered as representative of the UV-visible range, and the albedo at 865 nm is considered as representative of the near infrared. The gaseous absorption is estimated from TOMS data and from the POLDER 910- to 865-nm reflectance ratio. In a previous approach, the respective weights of the three narrowband albedos were based on radiative transfer simulations. Now, we take advantage of spatiotemporal coincidences between the second Advanced Earth Observing Satellite (ADEOS-2) and Terra to adjust these weights from the comparison between POLDER narrowband and CERES broadband reflectances. With no adjustment, the POLDER albedos are underestimated by 2% (in relative value) whereas the associated reflectances are underestimated by 4%. With an adjustment to the CERES reflectances, the POLDER shortwave albedos are overestimated by 2%. With or without adjustment the RMS difference between the POLDER and the CERES 1 degrees-regional instantaneous shortwave albedos is 6%; this is quite satisfactory since it is comparable to the CERES SW albedo consistency between nadir and oblique viewing zenith angles. When considering only homogeneous areas, the agreement between POLDER and CERES estimates was rather less good in the previous approach; it is now more than twice better. The difference between POLDER and CERES shortwave albedo estimates appears, for a large part, due to our plane-parallel assumption that is crucial for the heterogeneous liquid water clouds. It results in a bias in monthly mean shortwave flux around 2 W m(-2) that is found to present only small variations in latitude.