Simulation is used to evaluate the future performance of water supply headworks systems. Typically, most of the uncertainties associated with the use of demand models are ignored in the simulation. A simulation methodology is presented which makes allowance for the uncertainties associated with the use of demand and streamflow models. The demand uncertainties considered are natural, climatic, socioeconomic, and model parameter uncertainty. The methodology is illustrated using two case studies which consider a simple stationary demand model and the widely used trend extrapolation demand model. The results show that ignoring demand uncertainties (as well as streamflow model parameter uncertainty) can significantly bias the expected reliability. This bias depends on the magnitude of demand uncertainty which in turn depends on model structure, length of data available for calibration, and uncertainty in future socioeconomic variables. Moreover, this bias tends to increase as streamflow variability decreases. To guard against overoptimistic assessment of system performance, it is suggested that allowance be made for demand uncertainty (as well as streamflow model parameter uncertainty) in headworks simulation. ¿ American Geophysical Union 1993