An Overview of the CIRA Dense Optical Flow Retrieval and Applications from Rapid Scan Geostationary Satellite Imagery Project

Optical flow (OF) is defined as the distribution of apparent velocities of motion of brightness patterns in an image sequence. In geostationary satellite imagery, OF often is the result of cloud- and water-vapor-drift motions in the atmosphere which are related to the environmental winds. The Geostationary Operational Environmental Satellite (GOES)-R Advanced Baseline Imager (ABI) provides routine rapid scan satellite imagery which enables advanced techniques for dense (every image pixel) OF retrieval, which in turn enables several novel downstream products. This presentation will overview a 3-year NASA project to develop and compare retrieval approaches, including variational (or penalty function minimization) and machine-learning-based OF algorithms, which are provided on open-source code repositories, and identify how such techniques expand upon the state of the art used in operational atmospheric motion vectors. This presentation will also overview applications produced at the Cooperative Institute for Research in the Atmosphere (CIRA) as a result of this project, including a) the Speed and Direction sandwich products, which blend imagery with DOF rendered wind motions to highlight convection developing in strong vertical wind shear, b) cloud-top cooling products derived from optical flow warping of satellite imagery for nowcasting deep convection, c) cloud-top divergence, which highlights intensity trends of mature deep convection, d) MesoAnywhere, where DOF is used to infer 30-sec imagery using only 5-min refresh data, and e) cloud nowcasting, where DOF motions are used to estimate where cloud cover will exist up to 3 hours in the future. We will include discussions on common artifacts in these applications, and strategies learned for artifact mitigation. Finally, the presentation will discuss future research in optical flow retrieval, including efforts for DOF retrieval over lower temporal resolution imagery, which would greatly expand the spatial coverage of already developed optical flow products.