In this study, geological historical, and meteorological data were combined to produce a regional chronology of flood magnitude and frequency in nine small basins (7--70 km2). The chronology spans more than 1000 years and demonstrates that detailed records of flood magnitude and frequency can be compiled in arid regions with little to no conventional hydrologic information. The recent (i.e., post-1950) flood history was evaluated by comparing a 50-year series of aerial photographs with precipitation data, ages of flood-transported beer cans, anthropogenic horizons in flood sediments, postbomb 14C dates on flotsam, and anecdotal accounts. Stratigraphic analysis of paleoflood deposits extended the regional flood record in time, and associated flood magnitudes were determined by incorporating relict high-water evidence into a hydraulic model. The results reveal a general consistency among the magnitudes of the largest floods in the historical and the paleoflood records and indicate that the magnitudes and relative frequencies of actual large floods are at variance with 100-year flood magnitudes predicted by regional flood frequency models. This suggests that the predictive equations may not be appropriate for regulatory, management, or design purposes in the absence of additional, real data on flooding. Augmenting conventional approaches to regional flood magnitude and frequency analysis with real information derived from the alternative methods described here is a viable approach to improving assessments of regional flood characteristics in sparsely gaged desert areas. ¿ 2001 American Geophysical Union