How well can the albedos of complex cloud scenes be estimated ? One promising one- dimensional approximation to the full radiative transfer is the tilted independent pixel approximation (TIPA; Varnai and Davies, 1999). TIPA uses the distribution of optical thicknesses along the slant path of the Sun, so that at oblique Sun angles the cloud fraction as apparent to the Sun is used. TIPA is therefore more applicable to broken cloud scenes than IPA. TIPA s validity for realistic and complex cloud scenes, however, still needs to be established. An additional simplification, in which only the mean optical depth of a broken cloud is used, is also worth examining.

The question is explored using Terra satellite Multiangle Imaging SprectroRadiometer (MISR) 0.67 micron data. Cloud scenes are purposely chosen to be complex and broken. A necessary first step is to (three-dimensionally) model the cloud scene albedos as accurately as possible, including the clear-sky scene portions. This is done with a Monte Carlo model using as its lower boundary an analytical model (Breon, 1993) for ocean surface sun glint, aerosol, rayleigh and molecular scattering. The 275-m pixel cloud heights are derived using a stereo retrieval, and the cloud optical depths determined through an iteration between the modelled and observed reflectances.

Preliminary results show that the Monte Carlo modelled reflectances can match the observed reflectances of complex, broken cloud scenes quite well, with the largest discrepancies occurring at the most oblique viewing angles and within the sun glint region. Results for the relative accuracy of the TIP approximation will be presented at the conference.