BETTER ESTIMATING UNCERTAINTY IN GEOLOGIC PALEOFLOOD ANALYSES

Geologic-based paleoflood analyses have matured over the past decades and recently reached a major milestone with the inclusion of methods for including paleoflood data in flood-hazard frequency analyses (2017 Bulletin 17C update). These uses will eventually require more rigorous estimates of the uncertainties associated with geologic-based paleoflood data as the data are used more in formal hazard and risk analyses for high-value projects. For paleohydrologists, the experience and practice of the seismic hazard community (e.g. through SSHAC and UCERF3) offer lessons and a template for adaptation. This experience has shown that the views of multiple experts, careful identification of the aleatory and epistemic components, and tools such as logic trees may all have utility in better defining uncertainty associated with geologic paleoflood events used in flood hazard and risk analyses. At its most basic level geologic paleoflood information for flood hazards consists of three elements: 1) event identification, 2) age or timing, and 3) flood size (usually peak discharge); similar elements to those of individual paleoseismic events. Precedents for estimating and combining uncertainties of individual paleoseismic events across multiple exposures and dating methods are well evolved and readily adapted to paleoflood studies. For paleoflood analyses, addressing uncertainty related to Issues of stratigraphic completeness at individual sites, and between related sites, is key to developing confidence that a record of floods above some threshold is complete. The depth and velocity of flow associated with sediment deposition or erosion for a single flood will vary from site to site. Stratigraphic elevation of each site gives a minimum stage, but other stages may be estimated by independent reasoning, or output parameters from the flow models and empirical transport relations. Paleoflood discharge or stage estimates have uncertainties related to each of several flow model inputs and quasi-independent model assumptions. Uncertainty related to channel topography, geometry, roughness, vegetation effects, sediment transport, and model resolution can all have significant impacts. When combining results among different methods and sites, varied uncertainties, dependencies, and correlations may exist.