Hydrogen is likely to be the most important future energy carrier, for many stationary and mobile applications, with the potential to make significant reductions in greenhouse gas emissions especially if renewable primary energy sources are used to produce the hydrogen. A safe transition to the use of hydrogen by members of the general public requires that the safety issues associated with hydrogen applications have to be investigated and fully understood. In order to assess the risks associated with hydrogen applications, its behaviour in realistic accident scenarios has to be predicted, allowing mitigating measures to be developed where necessary. A key factor in this process is predicting the release, dispersion and combustion of hydrogen in appropriate scenarios. This paper illustrates an application of CFD methods to the simulation of an actual hydrogen explosion. The explosion occurred on 3 March 1983 in a built up area of central Stockholm, Sweden, after the accidental release of approximately 13.5 kg of hydrogen from a rack of 18 interconnected 50 l industrial pressure vessels (200 bar working pressure) being transported by a delivery truck. Modelling of the source term, dispersion and combustion were undertaken separately using three different numerical tools, due to the differences in physics and scales between the different phenomena. Results from the dispersion calculations together with the official accident report were used to identify a possible ignition source and estimate the time at which ignition could have occurred. Ignition was estimated to occur 10s after the start of the release, coinciding with the time at which the maximum flammable hydrogen mass and cloud volume were found to occur (4.5 kg and 600 m(3), respectively). The subsequent simulation of the combustion adopts initial conditions for mean flow and turbulence from the dispersion simulations, and calculates the development of a fireball. This provides physical values, e.g. maximum overpressure and far-field overpressure that may be used as a comparison with the known accident details to give an indication of the validity of the models. The simulation results are consistent with both the reported near-field damage to buildings and persons and with the far-field damage to windows. The work was undertaken as part of the European Integrated Hydrogen Project-Phase 2 (EIHP2) with partial funding from the European Commission via the Fifth Framework Programme.