Improvements to near-real-time cloud detection near the terminator and impact on diurnal trends of cloud properties
Poster PDF
Authors
Christopher Rogers Yost — Science Systems and Applications, Inc.
Patrick Minnis — NASA - Langley Research Center
—
Rabindra Palikonda — Science Systems and Applications. Inc./NASA - LRC
J Kirk Ayers — NASA - Langley Research Cntr/Science Systems and Application
Douglas A. Spangenberg — Science Systems and Applications, Inc.
Category
Cloud Properties
Description
Automated cloud detection near the day/night boundary known as the terminator is uniquely difficult because of the low signal-to-noise ratio of reflected solar radiation. Threshold-based cloud detection algorithms often have difficulty setting appropriate clear-sky thresholds, and consequently, cloud amount may differ significantly than that derived from daytime and nighttime methods. The NASA Langley Clouds and Radiation Group uses a modified version of the Clouds and Earth’s Radiant Energy System (CERES) cloud mask to identify clear and cloudy pixels in geostationary satellite imagery. Loops of cloud mask images derived from geostationary satellite imagery show that cloud identification near the terminator is often inconsistent with the preceding and succeeding images when the same region is processed with the daytime or nighttime methods. Cloud fraction is often overestimated near the terminator, and the magnitude is greater over land compared to water surfaces due to greater field of view complexity over land. Cloud physical and microphysical properties assigned to pixels mistakenly identified as cloudy contaminate studies of cloud life cycles and diurnal trends. This study presents a method, using the cloud fraction and clear and cloudy infrared brightness temperatures from previous satellite scans, to improve cloud detection accuracy near the terminator for a threshold-based cloud mask. The performance of the algorithm is evaluated using instrumentation at Atmospheric Research Measurement (ARM) Climate Research Facility sites such as the Southern Great Plains (SGP) site. The impact of false detections on hourly mean cloud properties is also investigated.
