Kotarba, AZ (2010). Estimation of fractional cloud cover for Moderate Resolution Imaging Spectroradiometer/Terra cloud mask classes with high-resolution over ocean ASTER observations. JOURNAL OF GEOPHYSICAL RESEARCH-ATMOSPHERES, 115, D22210.
Cloud masks are the final product of cloud detection and the starting point for determining the cloud amount. To calculate the cloud amount within thematic classes of cloud masks, the cloud fraction for each class must be assigned. This is usually done subjectively, assuming that clear pixels indicate a cloud fraction value of 0.0 and cloudy pixels indicate a cloud fraction value of 1.0. An analogous interpretation can also be applied using the Moderate Resolution Imaging Spectroradiometer (MODIS) level 3 approach (L3), which calculates global cloud amounts from instantaneous cloud masks. In this paper, the cloud fractions for each of the four MODIS cloud mask classes were derived in a more objective manner. They were based on high-resolution Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) observations that were collocated in space and time with MODIS data. This study was limited to cumulus and stratocumulus clouds observed during the daytime over the ocean surface with a MODIS nadir viewing geometry. Cloud fraction values were averaged and used to calculate the cloud amounts for all analyzed ASTER scenes, which were then compared to statistics that assumed an L3 approach. Our results show that the L3 approach tended to overestimate the cloud amount in general by 0.087, and this value ranged between 0.007 and 0.100, depending on the observation conditions. We found that cloud fraction values were influenced by the average cloud amount in the analyzed ASTER scenes, especially in the case of "cloudy" and "uncertain" pixels. The MODIS 250 m cloud mask was also evaluated and found to be a reasonable source of cloud fraction information for cloud mask classes during the daytime over the ocean, reducing L3 overestimation by approximately 30%.