Evaluating and Correcting the Effect of Sensor Spatial Resolution on Cloud Fraction Derived from Satellite Instruments

cloud fraction;remote sensing;resolution;pattern recognition

Jones, Alexandra L. ; Rauber, Robert M., ; ,Di Girolamo, Larry ; Di Girolamo ; Larry

关键词: cloud fraction;    remote sensing;    resolution;    pattern recognition;   

Others  :  https://www.ideals.illinois.edu/bitstream/handle/2142/18236/Jones_Alexandra.pdf?sequence=2&isAllowed=y

美国|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

This study examines the biases in cloud fraction (CF) derived from satellite instruments that are due to the effect of sensor spatial resolution. In total, 1405 15 m resolution ASTER (Advanced Spaceborne Thermal Emission and Reflection Radiometer) scenes over cumulus-dominated regions including the Indian Ocean, Caribbean Sea, and Gulf of Mexico were collected. Cloud masks of these scenes were treated as “truth” and used to study the dependence of CF biases on sensor resolution and cloud distributions. We found that at typical meteorological satellite resolution (~1 km) the traditional CF estimate derived from perfect clear-conservative masks has a median bias of ~0.28 in the above mentioned regions. To obtain a median bias on the order of 0.01, without correction, a cloud mask resolution less than 80 m is required. A simple equation can correct the bias to 0.025 at 150 m resolution and the more computationally expensive pattern recognition technique can correct the median bias to 0.0 at any resolution tested and for any true CF tested. We further applied the above correction techniques to an operational cloud mask, the RCCM (Radiometric Camera-by-camera Cloud Mask) of the MISR (Multi-angle Imaging Spectro-Radiometer) instrument, which has spatially and temporally overlapped observations with ASTER. We applied the pattern recognition technique to correct a CF climatology in the tropical Western Atlantic dominated by small trade wind cumulus. The average CF was reduced from 0.496 to 0.199. Although the results were simplified to the average improvement, this technique does more than simply subtract a standard value for bias to improve the climatology; the degree of correction for each 17.6 km x 17.6 km region of the larger cloud mask is dependent on the spatial distribution of clouds in that sub-scene.