A massive vapour cloud explosion occurred at the Buncefield fuel depot near Hemel Hempstead, UK, in the morning of 2005 December 11. The explosion was the result of an overflow from one of the storage tanks with the release of over 300 tons of petrol and generating a vapour cloud that spread over an area of 80 000 m2, before being ignited. Considerable damage was caused in the vicinity of the explosion and a total of 43 people were injured. The explosion was detected by seismograph stations in the UK and the Netherlands and by infrasound arrays in the Netherlands. We analysed the seismic recordings to determine the origin time of 06:01:31.45 ±0.5 s (UTC) from P-wave arrival times. Uncertainties in determination of origin time from acoustic arrival times alone were less than 10 s. Amplitudes of P-, Lg and primary acoustic waves were measured to derive decay relationships as function of distance. From the seismic amplitudes we estimated a yield of 2–10 tons equivalent to a buried explosion. Most seismic stations recorded primary and secondary acoustic waves. We used atmospheric ray tracing to identify the various travel paths, which depend on temperature and wind speed as function of altitude, leading to directional variation. Refracted waves were observed from the troposphere, stratosphere and thermosphere with a good match between observed and calculated traveltimes. The various wave types were also identified through array processing, which provides backazimuth and slowness, of recordings from an infrasound array in the Netherlands. The amplitude of stratospheric refracted acoustic waves recorded by the array microbarometers was used to estimate a yield of 21.6 (±5) tons TNT equivalent. We have demonstrated through joint seismo-acoustic analysis of the explosion that both the seismic velocity model and the atmospheric model are sufficient to explain the observed traveltimes.