These data describe the 10-largest annual, instantaneous flood peaks and corresponding dates of those largest floods for 473 streamgages in the Hydro-Climatic Data Network (HCDN) HCDN-2009 (Lins, 2012) network; they are also the data and results used in the companion manuscript by Collins, M.J, Hodgkins, G.A., Archfield, S.A., and Hirsch, R.M, under review, "The occurrence of large floods in the United States in the modern hydroclimate regime: seasonality, trends, and large-scale climate associations," and submitted to Water Resources Research. The HCDN-2009 (Lins, 2012) network is a subset of streamgages classified as unimpaired to the extent possible that any upstream regulation or urbanization could be evaluated. From this network, the set of streamgages was further narrowed based on their completeness of record and according to the criteria used in Hodgkins et al. (2019) and Dudley et al (2018), which requires that each decade of annual flood peaks have at least 8 observations or 80 percent completeness in each decade. Of the 473 streamgages, 386 streamgages overlapped with those streamgages classified by McCabe and Wolock (2014) into 14 geographically-coherent clusters that correspond to climatic and physiographic regions across the country and are noted in the dataset. These clusters were used to determine months with flood-rich or flood-poor months, also reported in the data release. For those streamgages not included in McCabe and Wolock (2014), streamgages were assigned to one of the 14 classes by applying the methods outlined in their paper. The metadata includes an attribute column to denote which streamgages fall into this category and which streamgages were included in McCabe and Wolock (2014).