This paper describes and demonstrates an approach to improve the management of risks from small-probability events that can lead to large consequences. It applies a decision-based theory to account for limited information in estimating frequencies for rare events to large rockfill dam in Norway that is being assessed for rehabilitation. Uncertainties are considered specifically in estimating the overtopping hazard for the existing dam and for an elevated dam crest. Uncertainty in the estimates of the overtopping hazard curve means that smaller costs of dam failure and/or larger costs of rehabilitation may be justified. From a practical perspective, a cost of rehabilitation in this case that is nearly ten times larger could be justified when the uncertainty in the estimate of the hazard curve is considered. The value of perfect information about the hazard curve increases as the amount of information available decreases and as the cost of failure relative to the cost of rehabilitation decreases. In this case, the value of perfect information about the hazard curve is about 25 percent of the cost to raise the dam crest