On 3 June 2004, a bolide was observed over British Columbia, Washington, Oregon, and Idaho. In addition to eyewitness accounts, the event was recorded on videotape and at ∼100 seismometers located in Washington State and southern British Columbia. Seismic records are consistent with a terminal burst rather than the hypersonic shock of the meteors passage through the atmosphere. Arrival times from seismic waveform data are used to obtain an accurate source location for this terminal burst. The source location can effectively be considered to be ground‐truth for assessing current atmospheric models and infrasound propagation algorithms. We observe clear infrasonic signals associated with the terminal burst of the fireball at infrasound stations I56US and I57US in Washington State and California, respectively. At I56US we observe at least four distinct acoustic arrivals, for which we are able to model three using a state‐of‐the‐art atmospheric model (the Ground‐to‐Space (G2S) model) and both ray‐tracing and parabolic equation propagation algorithms. To our knowledge, this is the first study in which a complex sequence of arrival packets at an individual station has been successfully modeled by range‐dependent ducting and specular reflection off terrain gradients. At I57US a single coherent acoustic signal was observed, which we are unable to satisfactorily model. This could be due to errors in the specification of the background atmosphere, incorrect assumptions and approximations inherent to the acoustic propagation algorithms, and the greater range of propagation, which amplifies the effect of errors in source location and the atmospheric specifications.