In Great Britain, advice on land-use planning decisions in the vicinity of major hazard sites and pipelines is provided to Local Planning Authorities by the Health and Safety Executive (HSE), based on quantified risk assessments of the risks to the public in the event of an accidental release. For potential exposures to toxic substances, the hazard and risk is estimated by HSE on the basis of a "toxic load". For carbon dioxide (CO2), this is calculated from the time-integral of the gas concentration to the power eight. As a consequence of this highly non-linear dependence of the toxic load on the concentration, turbulent concentration fluctuations that occur naturally in jets or plumes of CO2 may have a significant effect on the calculated hazard ranges. Most dispersion models used for QRA only provide estimates of the time- or ensemble-averaged concentrations. If only mean concentrations are used to calculate the toxic load, and the effects of concentration fluctuations are ignored, there is a danger that toxic loads and hence hazard ranges will be significantly under-estimated. This paper explores a simple and pragmatic modification to the calculation procedure for CO2 toxic load calculations. It involves the assumption that the concentration fluctuates by a factor of two with a prescribed square-wave variation over time. To assess the validity of this methodology, two simple characteristic flows are analysed: the free jet and the dense plume (or gravity current). In the former case, an empirical model is used to show that the factor-of-two approach provides conservative estimates of the hazard range. In the latter case, a survey of the literature indicates that there is at present insufficient information to come to any definite conclusions. Recommendations are provided for future work to investigate the concentration fluctuation behaviour in dense CO2 plumes. This includes further analysis of existing dense gas dispersion data, measurements of concentration fluctuations in ongoing large-scale CO2 release experiments, and numerical simulations. Crown Copyright (C) 2011 Published by Elsevier Ltd. All rights reserved.