Crown fires are complex, unstable phenomena dependent on feedback mechanisms between the combustion products of distinct fuel layers. We describe non-linear fire behaviour associated with crowning and the uncertainty they cause in fire behaviour predictions by running a semiphysical modelling system within a simple Monte Carlo simulation framework. The method was able to capture the dynamics of passive and active crown fire spread regimes, providing estimates of average rate of spread and the extent of crown fire activity. System outputs were evaluated against data collected from a wildfire that occurred in a radiata pine plantation in south-eastern Australia. The Monte Carlo method reduced prediction errors relative to the more commonly used deterministic modelling approach, and allowed a more complete description of the level of crown fire behaviour to expect. The method also provides uncertainty measures and probabilistic outputs, extending the range of questions that can be answered by fire behaviour models.