Using tsunami sediment transport experiments to improve paleohydraulic inverse models
Tsunami deposits can imperfectly record the hydraulic conditions of devastating extreme events. Sand entrainment, advection and deposition in these events occurs under strongly disequilibrium conditions in which traditional sediment transport models behave poorly. Quantitative models relating sediment characteristics to flow hydraulics hold the potential to improve coastal hazard assessments. However, data from recent natural tsunamis have rarely been accurate enough, over a range of parameter space, to quantitatively test proposed inverse models for predicting flow characteristics. To better understand how to “read” flow depth and velocity from disequilibrium deposits, we conducted controlled and repeatable laboratory flume experiments in which different grain size distributions (GSDs) of sand were entrained, transported and deposited by hydraulic bores. The bores were created by impounding and instantaneously releasing ~6 m^3 of water with a computer-controlled lift gate. The experiments represent 1/10 to 1/100 scale physical models of large events. Both flow characteristics (including Froude numbers) and suspended sediment transport characteristics (including Rouse numbers and grain size trends) scale consistently with documented natural tsunamis.
We use the experimental data to interpret how entrainment, transport and mixing influence deposit GSDs along the flume. Suspension-dominated deposits get finer and thinner in the direction of transport. The data show that two key controls on GSDs along the flume are (a) the size distribution of the sediment source, and (b) turbulent dispersion of grains. First, the influence of source GSDs on deposit GSDs is strongest near the sediment source. Size-dependent suspension and settling become increasingly important farther down the flume. Transport distances of 1-2 advection length scales are required for deposit GSDs to be sensitive to flow hydraulics. Second, turbulent dispersion strongly influences local deposit GSDs. Importantly, intermediate deposit grain size percentiles (e.g. D50) are less sensitive to dispersive transport than either the fine or coarse tails of local deposit GSDs. Using deposit GSDs along the flume, an advection-settling model best predicts flow depths and velocities when calculated for intermediate percentiles (e.g. D50), rather than for coarse size fractions (e.g. D95) as has been assumed in previous works. We also highlight areas where our knowledge and predictive ability is limited and could be improved using experiments, including understanding the degree to which grain size sorting occurs during entrainment into suspension, and also during energetic bedload transport. Overall, the work suggests that physical models of tsunami sediment transport and deposition are invaluable for evaluating equation assumptions, benchmarking model results, and rigorously evaluating model uncertainties.