Environmental context Throughout the world there are many places where ozone levels are elevated above internationally accepted guidelines set to protect human health. Policy makers use air quality models to formulate emission control strategies to achieve these air quality goals for ozone. There are large uncertainties in these air quality models that mask the sensitivity of the model control strategies to chemical mechanism choice. Abstract Monte Carlo sampling of pre-specified parameter ranges has been used to replace a single ‘best estimate’ photochemical trajectory model run with 11 694 ‘acceptable’ model runs that are each consistent with the observations of elevated O3 during the PUMA (Pollution in the Urban Midlands Atmosphere) campaign in the UK, West Midlands during 1999. These 11 694 ‘acceptable’ parameter sets were then used for probabilistic evaluation of photochemical oxidant control strategies, based on 30% reductions in volatile organic compounds and NOx precursor emissions and on precursor emission projections to 2020. The sensitivity of single ‘best estimate’ model runs to chemical mechanism choice gave some indication of the robustness of photochemical oxidant control strategies. However, Monte Carlo parametric uncertainty analysis showed that sensitivity to mechanism choice failed to indicate the magnitudes of the likely uncertainty ranges in the O3 responses to photochemical oxidant control strategies. Furthermore, Monte Carlo uncertainty analysis showed that there may be O3 air quality disbenefits from 30% NOx emission reduction that were not apparent from ‘best estimate’ runs.