Dense optical wind flows


Jason Apke presents a method to derive winds from satellite data and shows some example applications.


The new Geostationary Operational Environmental Satellite (GOES)-R Series Advanced Baseline Imager and similar instruments represent a substantial improvement in the spatial, spectral, temporal, and radiometric resolution of operational space-based imagery for atmospheric sciences. The enhanced resolutions enable the objective derivation of fine-scale brightness motion, or so-called "optical flow" (OF), over a variety of meteorological phenomena with techniques that address the weaknesses of current generation cloud and feature tracking algorithms. The techniques even allow for precise dense OF derivation, or motion retrieval at every image pixel, which has a variety of applications that will be invaluable to satellite remote sensing-based forecasting and research in the future. These applications include Atmospheric Motion Vector fields, temporal brightness interpolation, feature extrapolation, image stereoscopy, and semi-Lagrangian brightness temperature field derivation. As motion is an observable that provides unique context on features observed with an image, OF fields further offer significant new predictors to a variety of objective machine learning and decision-making tools now in development. This presentation will go into detail on how some of these new dense OF techniques are derived and highlight efforts to produce and validate new products. Demonstrations of novel RGBs and their uses for operational forecasting will be overviewed, such as blends of derived speed for inferring shear and cloud-top cooling for inferring vertical growth with clouds in visible satellite imagery. Efforts to validate wind products will also be shown which highlight the strengths and weaknesses of dense OF derivation for satellite remote sensing purposes. Examples of future applications and research will also be covered, including a look into the future of OF derivation with an example 6-sec GOES-17 dataset.

Go to Webcast...

Lecture slides...