Advancements in Artificial Intelligence and Machine Learning to Improve Detection and Classification of GOES-R Ground System Product Artifacts

Quality Watch is an automated AI/ML ground processing artifact detector and classifier trained to operate on ABI imagery. Its primary purpose is detecting ground system issues that affect product generation. Training data was generated for missing blocks, Caterpillar Tracks, Solar Avoidance, and Shark Fin artifacts using ABI L1b Cloud and Moisture Imagery available from NOAA’s Big Data project. The model has been migrated to TensorFlow, optimized and hosted in an AWS cloud instance, processing GOES-16 ABI data in near real time. The use of TensorFlow combined with model enhancements has resulted in significant performance improvements to runtime, along with reductions in resource utilization. Quality Watch provides ground system operators and engineers a cost-effective, automated means of monitoring GOES-R GS Data Operations at improved pace and scale.