Cloud Nowcasting for Developing and Dissipating Clouds

The Optical Variability Evaluation of Regional Cloud Asymmetries in Space and Time (OVERCAST) project aims to identify and forecast three-dimensional (3D) cloud structures globally in the next 0-3-hours. An integral part of the OVERCAST project is short-term cloud forecasting, commonly referred to as nowcasting. Cloud nowcasting efforts support operational applications, including Department of Defense operations, severe weather forecasting, solar power generation, and aviation, by providing accurate and timely indications of cloud presence in the immediate future. While OVERCAST’s leading optical flow-based nowcasting methods perform well for existing clouds, they are less effective at predicting the formation and dissipation of clouds. To address this gap, this project uses deep learning, a subset of machine-learning, to enhance OVERCAST’s existing cloud nowcasting approach. To identify and predict the development of deep convective clouds the deep learning models use satellite imagery, anvil detection techniques, and time series data. To accurately nowcast low-level cloud dissipation, satellite imagery optical flow retrievals are used to create a machine-learning labeled dataset of cloud dissipation timeframes. This presentation provides an overview of these efforts led by the Cooperative Institute for Research in the Atmosphere (CIRA).