GOES and Himawari Imagery Visualization Using GPUs for Full Scale Animation: Deriving 4-d Training Data for AI

Multi gigabyte Advanced Baseline Imager (ABI) and Advanced Himawari Imager (AHI) scenes are generated many times a day from GOES 16, presumably GOES 17 once operational, as well as Himawari-8. Visual interrogation of these data can be challenging due to the large size of the imagery and the collection frequency. Harris Geospatial Solutions developed methods to push full scale resolution ABI three band composites or single band time series to an NVIDIA Graphical Processing Unit. By using the graphics card for rapid visualization, ABI data can be animated at full resolution, panned, zoomed in/out, annotated and regions of interest created while the data are animating. This kind of interactive visualization enables points or areas of interest to be tracked across time and motion. Certain weather phenomenology such as tornadic thunderstorms, wind and/or rain events and other weather activities can be combined with other datasets like lightening data from GLM as an overlay. While the data products are animating, areas of interest can be drawn by hand or if certain thresholds exist, extracted automatically. The resulting areas then represent geographic regions that have spectral, spatial and temporal characteristics. These training data regions can then be further analyzed, fused or combined with other datasets to provide input into AI algorithms to better understand and predict weather patterns. This presentation will discuss how rapid visualization for large data sets can be built, show examples, and discuss the use of these tools to impact analytics as well as communicate satellite observed weather phenomenology to the scientific community.