Finding Fireballs Hidden in Satellite Lightning Data
Weather satellite data contains a wealth of information well beyond applications to meteorology. The National Ocean and Atmospheric Administration (NOAA) operates the Geostationary Operational Environmental Satellite’s (GOES) Geostationary Lightning Mapper (GLM) instruments, which detect millions of lightning strikes per day using a near-infrared optical sensor that can detect the momentary changes in light. However, a handful of these events aren’t lightning at all, but bolides—large exploding meteors in Earth’s atmosphere. Funded by NASA's Planetary Defense Coordination Office, researchers at the SETI Institute are utilizing systems at the NASA Advanced Supercomputing (NAS) facility to develop an automated pipeline to identify this small number of bolides. The goal is to create a rich, calibrated, and statistically consistent dataset of bolide light curves to inform the planetary defense community of the risks associated with meteoroid and large asteroid impacts.
Quick Facts
We are truly trying to find needles in a haystack. Within the millions of lightning strikes and hundreds of terabytes of data, there are a handful of interesting bolides. The supercomputers and fast Lustre filesystems at the NASA Advanced Supercomputing facility provided the massively parallel resources to reprocess all historical data in a matter of hours. This has resulted in rapid prototyping and development of the detection pipeline.
- The custom pipeline utilizes Python-based machine learning tools, and reprocessing the full GOES GLM dataset requires massive distribution over NASA’s Aitken supercomputer. Additionally, training machine learning models also utilizes GPU nodes provided by NAS.
- The team is detecting upwards of dozens of bolides per day, where it was originally believed about one per week would be identified. After over four years of processing, this unexpected dataset is now large enough to perform statistical modeling of the distribution of large meteoroid impacts over the Earth.
- Their goal is to create a large, calibrated dataset that can be used to tune entry modelling software for incoming large asteroids to inform the planetary defense community of the potential risks of impacts.
- Eventually, the researchers aim to have a fully automated pipeline with no human intervention that would automatically post bolide detections online quickly.
Researchers
- Jeffrey C. Smith, SETI Institute Please enable Javascript to see email address.
- Nina McCurdy, NASA Ames Research Center Please enable Javascript to see email address.
