OBJECTIVE: To model the possible air pollution-related health impact of the 2005 oil depot fire at Buncefield, near London, UK, under alternative meteorological conditions to those experienced at the time. DESIGN: Atmospheric dispersion modelling of the smoke plume was conducted under the range of meteorological conditions occurring throughout 2005 assuming constant particle emission rates. Population exposure to particle concentrations (PM10) was calculated by linking the atmospheric dispersion modelling data (2 km resolution) and postcode population data. Health impacts were estimated using time-series-based exposure-response relationships for PM10 available from the epidemiological literature. MAIN OUTCOMES: Estimates of pollution-related deaths brought forward, emergency hospital admissions from respiratory problems and emergency hospital admissions from cardiovascular disease. FINDINGS: The highest four-day population exposure to PM10 for meteorological data from 2005 was predicted to occur between 5 and 8 August 2005, when northerly winds would have carried the plume towards London and surrounding areas of high population density. On these days, we estimated the additional PM10 exposure would have resulted in around 12 extra deaths brought forward, and around 13 additional emergency hospital admissions and a similar additional number of emergency admissions for cardiovascular disease. These numbers are slightly greater than estimated deaths and emergency admissions attributable to regular anthropogenic PM10 concentrations in south east England over the same four day period. CONCLUSIONS: Although the particle pollution-related health impacts of the Buncefield fire could have been higher under different meteorological conditions, it is unlikely that the impacts would be substantially greater than those attributable to regular anthropogenic particle pollution over the similar period. KEYWORDS: oil depot fire; health impact; epidemiology; air pollution; explosion; atmospheric dispersion modelling; exposure.