Seismoacoustic recovery of the SpaceX CRS-32 Dragon re-entry with a proof-of-concept for automated detection

Mikel Taotao Yu , Benjamin Fernando 

Atmospheric re-entries of human-made objects are increasing in frequency, yet detailed information on their trajectory, fragmentation, and debris footprint is rarely made publicly available. This creates an information asymmetry in which regulators, independent agencies, and the scientific community lack the means to verify that re-entries occurred as reported or to characterize their physical processes for future modeling. We address this gap by applying a seismoacoustic localization method to the 25 May 2025 re-entry of the SpaceX CRS-32 Cargo Dragon spacecraft which splashed down in the Pacific Ocean off the coast of Southern California. Using openly available seismic waveform data from 5 networks, including 22 Raspberry Shake amateur instruments, we recover a re-entry trajectory consistent in trajectory with the operator's publicly communicated splashdown prediction and release the manually labeled CRS-32 labeled N-wave picks as a public dataset. As a preliminary step toward automation, we train a neural network on the manually picked dataset and demonstrate reliable onset detection with stable performance across a wide range of detection thresholds, showcasing promise for N-wave detection from a two-event training set. These results show that existing open seismic infrastructure can provide independent, continental-scale characterization of re-entry events: a capability increasingly needed as the rate and opacity of commercial re-entries continue to grow.