Cloud assessment for ASTER nighttime scenes is not accurate because the ASTER Cloud Coverage Assessment
Algorithm (ACCAA) thresholds with only one thermal infrared (TIR) band for nighttime scenes. First in the present
paper, it is shown that the original ACCAA cloud masks differ considerably from the masks interpolated from MODIS
Cloud Mask Products (MOD35), and this discrepancy is caused from errors in the ACCAA masks by visual check for
543 scenes. In addition, uncertain pixels included in MOD35 masks, which are classified to neither cloud nor clear, are
visually checked for 76 scenes. Then, the ASTER nighttime cloud mask database using MOD35 products is introduced.
It provides the interpolated MOD35 cloud masks for almost all ASTER nighttime scenes (143,242 scenes as of July
2008) through Internet. The database also shows that clear scenes with cloud coverage of 20% or less are about 34% of
the total nighttime scenes. In the final part of the paper, an algorithm for reclassifying an interpolated MOD35 mask
using ASTER measurements is proposed and applied to 42 test scenes. The algorithm will work well for some scenes,
but less well for snow/ice surfaces, and thin, cirrus, and high clouds, due to the band limitation of ASTER/TIR. If a
spatial uniformity test is added, the algorithm performance may be improved.