The project investigates a new method for estimating the exceedance probabilities of extreme floods. The method is based on separately estimating the probability distribution of flood volumes and the distribution of flood peaks conditioned on volume. Preliminary results from the upper midwestern United States indicate that flood volumes are very well behaved compared to peaks, and suggests that regionalized estimates of their probability can be extrapolated more credibly than can the estimates of probability of flood peaks. These results will be tested using flood data from other regions of the United States as well. Various methods of regionalizing the distribution of flood volumes also will be explored. In estimating the distribution of flood peaks conditioned on volume, the deterministic watershed modeling will be used in the formulation and estimation of the statistical model. This approach will be developed and applied to one or more drainage basins. The watershed model will incorporate a one-dimensional, full dynamic flood routing model in order to capture the impact of floodplain storage on flood wave propagation. The results of this work will improve the understanding of basic rainfall-runoff-flood probability relationships, and will result in a significant reappraisal of techniques used in risk analysis of major engineering water projects and existing structures.