The data quality (third and final) extension of the data cubes seems to contain only integers (0, 1, ..., 6 in a cube I just looked at), however the array is stored as a double-precision float (8 bytes per value) instead of a 1-byte integer. This means that every cube file could be (7/8)*(1/3) = 30% smaller.
The data quality (third and final) extension of the data cubes seems to contain only integers (0, 1, ..., 6 in a cube I just looked at), however the array is stored as a double-precision float (8 bytes per value) instead of a 1-byte integer. This means that every cube file could be (7/8)*(1/3) = 30% smaller.