A008-0013
Detection of Low Cloud in Multilayer Scenes with the GOES ABI

Monday, 7 December 2020
Poster
John M Haynes1, Yoo-Jeong NOH1, Steven D Miller2 and Andrew Heidinger3, (1)Cooperative Institute for Research in the Atmosphere, Colorado State University, Fort Collins, CO, United States, (2)Colorado State University, Cooperative Institute for Research in the Atmosphere, Fort Collins, CO, United States, (3)NOAA National Environmental Satellite, Data, and Information Service, Silver Spring, MD, United States
Abstract:
The newest generation of geostationary sensors, including the GOES Advanced Baseline Imager (ABI), have provided new opportunities for remote sensing of clouds in the Earth's atmosphere. Detection problems associated with multiple-layered cloud systems, however, remain. In particular, it is difficult to infer the boundaries of cloud layers from passive sensor measurements alone, and multiple cloud layers makes this problem more difficult. Yet some satellite customers, particularly in the aviation community, would benefit from products that allow detection of low clouds underneath higher cloud layers.

To enhance information about the vertical extent of cloud layers, we have developed a statistical algorithm that derives the base of the upper-most cloud layer given cloud top height and cloud water path, based on statistical relationships obtained from MODIS and CloudSat/CALIPSO data. This information has been used to enhance the ABI Cloud Cover Layers (CCL) product that identifies cloud occurrence in vertical layers, but it still has limitation in multilayer cloud scenes.

In this study, an analysis of the improved Cloud Cover Layers algorithm will first be presented. Then we will describe development of a machine learning (Random Forest) model designed to place clouds vertically in a column using a priori observations from CloudSat and CALIPSO. The combination of these observations with estimates of the vertical profile of moisture in the atmosphere has allowed development of an ABI-based algorithm for low-cloud detection (which is also applicable to other sensors, including VIIRS). Using machine learning, the probability of detection of low clouds within ABI-identified cloudy scenes is approximately 90%, with a false alarm ratio of under 15%. In particular, when such clouds occur under cirrus (one of the most difficult retrieval situations), our method increases the probability of detection of such clouds more than threefold, from 21% to 71% (with a false alarm ratio of 21%). Finally, we will illustrate application of this method to ABI data, demonstrating how it presents a more complete picture of the vertical distribution of cloudiness on Earth, and describe some applications.

<< Previous Abstract | Next Abstract >>