In general, odours are perceived by humans through a single respiratory act having a duration of approximately 5 seconds. Moreover, odour concentration varies almost randomly over time, as any other physical variable in a turbulent environment. For this reason, both field measurements and dispersion models used for odour impact assessment purposes should be able to refer to short sampling times, instead of considering concentrations averaged over prolonged periods (i.e. 1 hour). This paper has the aim to investigate the problem of odour concentration fluctuations and their correlation with odour nuisance, thereby briefly discussing the problem of turbulence based on some theoretical considerations and of a few experimental results. The experimental results relevant to the continuous monitoring of instantaneous H2S concentration in ambient air clearly show that peak concentration exceed 1-hour average concentrations by almost 1 order of magnitude. Finally, the paper discusses the possible implementation of a dispersion model specifically dedicated to odour nuisance evaluation, thereby investigating different solutions to account for concentration fluctuations. The investigation related to the peak-to-mean approach, currently being the most widely used method to evaluate peak concentrations from 1-hour averaged concentrations on a regulatory level, allowed to point out the drawbacks of this methodology, which, despite its simplicity and ease-of-use, is unable to accurately describe the concentration fluctuations phenomena.