The following figure presents the differences between the "directional" and the angulary-averaged cloud spherical albedo derived from POLDER. Cloud thermodynamic phase discrimination is based on physical principles presented hereafter.
Here, clouds are assumed to be homogeneous plane-parallel layers composed of liquid water droplets (effective radius 10 µm), whatever the thermodynamic phase. If the microphysical model was well-adapted, the retrieved cloud spherical albedo would be the same in all the directions (see Doutriaux-Boucher et al, GRL, 2000, Parol et al, ASR, 2004).

What's happened when all clouds are assumed to be composed of spherical droplets ?

The discrepancy observed for ice clouds decreases when the Inhomogeneous Hexagonal Monocrystal model (IHM) is used

Various cloud particle models have been tested using POLDER1 data. One of the best fits is obtained using the IHM model. This model corresponds to randomly oriented hexagonal ice crystals containing air bubbles (C.-Labonnote et al. GRL, 2000; JGR, 2001). In the second generation of POLDER "ERB & cloud" processing line, this ice crystal model is used for the calculation of optical depth and albedo in the case of ice clouds

The following figure presents the comparison between the ice cloud optical thickness retrieved by using the ice crystal model and the liquid water droplet model as in Buriez et al. (IJRS, 1997). The comparison is restricted to the pixels observed from ADEOS1/POLDER on June 25, 1997 and totally labelled as "ice" by our cloud phase algorithm. The full line is slope unity. The short- and long-dashed lines correspond to the mean slope respectively for the marine and for the continental situations. The mean ratio between ice-model derived and liquid water-model derived optical thicknesses is 1.65 for oceanic ice clouds at 865 nm and 1.73 for continentals ice clouds at 670 nm (Buriez et al, JGR, 2004).

The linear mean cloud optical thickness at 870 nm over ocean and 670 nm over land results from an averaging over cloudy pixels and a weighting over available directions. An example of such a result is presented in the following figure for June 25, 2003. The spatial distribution of the cloud optical thickness is obviously coherent with the cloud structures observed in the cloud cover global map on the same day (see here-above): high mean values all along the ITCZ, over midlatitudes low-pressure cloudy systems and subtropical zones of low level clouds west of continents.