Australia experienced its first ever outbreak of equine influenza in 2007, affecting around 70,000 horses before eradication was achieved, 5 months later. We developed a spatial simulation model of equine influenza in Australia parameterised by Bayesian inference of data from the 2007 outbreak. Our modelling objectives were to improve understanding of the underlying disease process and the effectiveness of certain control measures, and to develop a real-time decision-support tool for future outbreaks. Local spatial spread, the movement of infected horses and intra-premises transmission were explicitly modelled. A Bayesian reversible jump Markov chain Monte Carlo algorithm was used to infer posterior distributions of key epidemiological parameters for a peri-urban and a rural cluster (n=7,670 and 1,540 horse premises, respectively). We then conducted forward simulation, a sensitivity analysis and validated against a previous simulation model of the 2007 outbreak. The simulated epidemics were highly cross-correlated both temporally and spatially with the actual epidemic (Spearman's ?=0.81 and 0.88, respectively) and with outputs of the previous model (?=0.84 and 0.92). Spread dynamics differed markedly in the two clusters, with local spread occurring over longer distances but at a lower rate in the rural cluster. The duration of infectivity was established for premises holding a range of numbers of horses and the effectiveness of vaccination was assessed. Our posterior parameter distributions will provide informative priors for future real-time modelling of equine influenza outbreaks in Australia, and our findings suggest which outbreak control procedures to prioritise. This research was funded by the RIRDC and the ABCRC. (Texte intégral)