[1] Epistemic uncertainty is a result of knowledge deficiency about the system. Sampling error exists when limited amounts of hydrologic data are used to estimate a T year event quantile. Both the natural randomness of hydrologic data and the sampling error in design quantile estimation contribute to the uncertainty in flood damage estimation. This paper presents a framework for evaluating a flood-damage-mitigation project in which both the hydrologic randomness and epistemic uncertainty due to sampling error are considered in flood damage estimation. Different risk-based decision-making criteria are used to evaluate project merits based on the mean, standard deviation, and probability distribution of the project net benefits. The results show that the uncertainty of the project net benefits is quite significant. Ignoring the data sampling error will underestimate the potential risk of each project. It can be clearly shown that adding data to existing sample observations leads to improved quality of information, enhanced reliability of the estimators, and reduced sampling error and uncertainty in the project net benefits. Through the proposed framework, the proper length of the extended record for risk reduction can be determined to achieve the required level of acceptable risk.