Vertical accretion sand proxies of gaged floods along the upper Little Tennessee River, Blue Ridge Mountains, USA

Understanding environmental hazards presented by river flooding has been enhanced by paleoflood analysis, which uses sedimentary records to document floods beyond historical records. Bottomland overbank deposits (e.g., natural levees, floodbasins, meander scars, low terraces) have the potential as continuous paleoflood archives of flood frequency and magnitude, but they have been under-utilized because of uncertainty about their ability to derive flood magnitude estimates. The purpose of this paper is to provide a case study that illuminates tremendous potential of bottomland overbank sediments as reliable proxies of both flood frequency and magnitude. Methods involve correlation of particle-size measurements of the coarse tail of overbank deposits (> 0.25 mm sand) from three separate sites with historical flood discharge records for the upper Little Tennessee River in the Blue Ridge Mountains of the southeastern United States. Results show that essentially all floods larger than a 20% probability event can be detected by the coarse tail of particle-size distributions, especially if the temporal resolution of sampling is annual or sub-annual. Coarser temporal resolution (1.0 to 2.5 year sample intervals) provides an adequate record of large floods, but is unable to discriminate individual floods separated by only one to three years. Measurements of > 0.25 mm sand that are normalized against a smoothed trend line through the down-column data produce highly significant correlations (R2 values of 0.50 to 0.60 with p-values of 0.004 to < 0.001) between sand peak values and flood peak discharges, indicating that flood magnitude can be reliably estimated. In summary, bottomland overbank deposits can provide excellent continuous records of paleofloods when the following conditions are met: 1) Stable depositional sites should be chosen; 2) Analysis should concentrate on the coarse tails of particle-size distributions; 3) Sampling of sediment intervals should achieve annual or better resolution; 4) Time-series data of particle-size should be detrended to minimize variation from dynamic aspects of fluvial sedimentation that are not related to flood magnitude; and 5) Multiple sites should be chosen to allow for replication of findings.