A large explosion occurred at the Buncefield Oil Depot, United Kingdom, on the 2005 December 11, as the result of a spillage of 300 tonnes of petrol. The explosion generated large amplitude infrasonic signals that were recorded at ranges of up to 1400 km. We present an analysis of data from across Central Europe using seven infrasound arrays, four of which recorded a total of 20 infrasonic arrivals, and 49 seismometers, which recorded 89 air-to-ground coupled arrivals. Exceptionally high amplitude stratospheric winds occurred at the time of the explosion, propagating east–southeast with speeds of up to 130 m s−1, generating a highly efficient duct for infrasound propagation. The Buncefield explosion can therefore be considered as a benchmark for studying the influence of different propagation algorithms and atmospheric models on phase identification, source location and yield estimation. Significant improvements are made in phase identification using accurate atmospheric parametrizations and 3-D modelling, where we identify 18 of the 20 observed phases, compared to seasonal atmospheric models and 1-D ray tracing which only identify seven of the phases. Additionally, various location scenarios are considered, including cross-bearings and least-squares traveltime and backazimuth inversion, all of which estimate locations within 35 km of ground truth. A novel single array location method using multiple phases works comparably well. Yield estimates made using the stratospheric arrival amplitudes exhibit one order of magnitude variability across the network, with a mean estimate of 51 tonnes high explosive equivalent. Detailed studies, such as this, are important for developing a successful monitoring regime for atmospheric or surface events, and for assessing the capability of the infrasound monitoring component of the Comprehensive Nuclear-Test-Ban Treaty verification system.