When weather events are expected to produce significant delays at destination airports, the traffic flow management system in the United States holds departing aircraft on the ground in an attempt to reduce delay costs to the operator and to alleviate airborne congestion. Selecting the correct amount of ground holding is made difficult because of uncertainty in predicting weather events that produce congestion. In general, decision-makers must strike a balance between the amount of predicted delay absorbed on the ground and the amount absorbed in the air. This paper first addresses the question of how uncertainty in the required delay should influence the amount of ground holding. It then establishes the relationship between delay uncertainty and uncertainties in predicting the onset and duration of weather events. Delay costs are minimized under an assumption that there is a fixed ratio between the cost of a unit of ground delay and a unit of airborne delay and that the landing sequence employed at the destination terminal is based upon the originally scheduled landing order. The analysis indicates that uncertainty in the delay prediction must be considered in selecting the optimum amount of ground holding for an individual flight. In predicting delays, it is desirable to keep the ratio of the standard error to the mean delay (σ / μ) well below 1.0 in order to avoid loss of benefits. A corresponding figure of merit for weather systems is shown to be the ratio of the uncertainty in onset/termination times to the duration of the weather event. Weather prediction systems must keep this ratio well below one-third to avoid significant loss of ground-holding benefits. The analysis indicates that reductions in the delay uncertainty through improved weather forecasting and traffic management systems can result in better decision-making and significant cost savings.