An Advanced Cloud Detection Approach for Creating Diurnally Consistent Geostationary Satellite Cloud Climate Data Records

Geostationary satellites (GEOsats) provide continuous cloud and meteorological observations over fixed portions of the Earth’s surface, allowing them to monitor the development and movement of storm systems and their diurnal variation. For climate studies, geostationary observations provide added insight into cloud formation and evolution and how they influence the diurnal cycle of Earth’s radiation budget.

A long and consistent cloud record can be a valuable resource for evaluating changes in cloud systems and properties across the globe. Integrating observations from different geostationary instruments poses challenges due to their distinct characteristics, such as different spectral channels and calibrations as well as varying spatial resolutions to list a few.  As a result, deriving consistent cloud properties from multiple sensors without introducing artificial discontinuities in a time series remains a complex and challenging endeavor.

A homogenized GEOsats cloud retrieval system is being developed to create cloud climate data records (CDR’s) for NASA’s CERES (Clouds and the Earth’s Radiant Energy System) mission from a long record of GEOsats that uses spectral channels common to most satellites. Thus, a 3-channel (0.6, 3.9, 11 µm) algorithm for daytime cloud detection, and a 2-channel (3.9 and 11 µm) algorithm for nighttime have been implemented and tested. Recent advances to the 3-channel processing framework include refined radiative transfer models specific to each GEOsats’ spectral bands to provide more accurate and consistent computed clear-sky TOA radiances. Machine learning approaches are also developed and implemented for estimating the a priori land surface skin temperature, and to improve cloud detection in the solar terminator and in oceanic areas with sunglint. It is anticipated that these changes will lead to more accurate and diurnally consistent derived cloud properties across satellite platforms.

This paper describes the CERES 3-channel cloud detection approach and presents results of initial cross-platform consistency and accuracy tests and evaluations with independent data from active sensors, such as CALIOP data, as well as from GEOsats analyses that utilize more spectral information. Remaining challenges and future work will be discussed.