Detecting diurnal cycle and lifetime of pyrocumulonimbus using GOES-16 infrared data with a machine learning model 

Satellite infrared (IR) cloud imagery has proven valuable in the identification of Pyrocumulonimbus (pyroCb) clouds. The substantial brightness temperature difference observed between warm shortwave IR wavelengths (~4 μm) and window IR wavelengths (~11 μm) has served as a reliable marker for detecting daytime pyroCb. However, this indicator becomes ineffective during nocturnal hours when the enhanced brightness temperature at 4 μm is solely a daytime phenomenon, arising from PyroCb microphysics that increase solar reflectivity of clouds. We have developed a machine learning model designed to detect pyroCb events during nighttime using IR channels from the Advanced Baseline Imager (ABI) aboard GOES-16. The model leverages the distinctive characteristics of daytime IR channels as its training data. We applied the trained model to five intense pyroCb events in western North America during August 2017. Furthermore, we have employed an established cloud-tracking tool known as Tracking and Object-Based Analysis of Clouds (tobac) to analyze the evolution of the clouds plumes and infer their lifetimes. Our research aims to extend this case study on a global scale, with the objective of creating a comprehensive database for the lifetimes of pyroCb events. Such a database will enhance our understanding of pyroCb dynamics, which is helpful for investigating the radiative implications and the potential impact on stratospheric chemistry.