Premium
Impact of Moderate Resolution Imaging Spectroradiometer (MODIS) cloud mask interpretation on cloud amount estimation
Author(s) -
Kotarba Andrzej Z.
Publication year - 2015
Publication title -
journal of geophysical research: atmospheres
Language(s) - English
Resource type - Journals
eISSN - 2169-8996
pISSN - 2169-897X
DOI - 10.1002/2015jd023277
Subject(s) - moderate resolution imaging spectroradiometer , cloud computing , cloud top , remote sensing , spectroradiometer , environmental science , cloud height , cloud cover , cloud fraction , satellite , meteorology , computer science , range (aeronautics) , mean squared error , statistics , mathematics , geography , physics , optics , reflectivity , astronomy , operating system , materials science , composite material
Cloud masks serve as a basis for estimates of cloud amount, which is an essential parameter for studying the Earth's radiation budget. The most commonly used cloud mask is a simple thematic classification, which includes qualitative information on the presence of clouds in the satellite's instantaneous field of view (IFOV). Cloud mask classes have to be “translated” into a quantitative measure, in order to be used for cloud amount calculations. The assignment of cloud fractions to cloud mask classes is a subjective process and increases uncertainty in cloud amount estimates. We evaluated this degree of uncertainty using the Moderate Resolution Imaging Spectroradiometer (MODIS) cloud mask product. Together with the operational MODIS cloud mask interpretation, we investigated two extreme alternatives: “rigorous” (only “confident cloudy” IFOVs were 100% cloudy) and “tolerant” (only “confident clear” IFOVs were 0% cloudy). Results showed that the range of uncertainty was 14.3% in Europe and controlled by the frequency of small convective clouds. Comparison with surface‐based observations suggests that the rigorous interpretation of the cloud mask is more accurate than that used operationally for MODIS level 3 product generation. The rigorous approach resulted in the smallest bias (−0.7%), the smallest root‐mean‐square error (4.6%), the small standard deviation (6%), and the strongest correlation (0.935). These results suggest that for climatological applications the rigorous scenario should be considered as a more accurate “best guess” over land.