A field and numerical study exploring uncertainty in high-fidelity modeling of riverine flows

Hydrodynamic modeling of natural rivers using large-eddy simulations (LESs) can provide researchers with valuable insights and detailed estimates of the flow field.  However, uncertainty inherent in model input parameters produce uncertainty in the model results.  This study collected field data to estimate uncertainty in the flow rate and bed roughness to complete a suite of LESs quantifying the propagation of uncertainty in the simulation results. The study area includes  a 0.6 km reach of the Sacramento River, California. We employed the polynomial chaos expansion (PCE) method to develop a surrogate model of the flow data for the streamwise velocities and vertical profile at select locations of the river. The PCE results were randomly sampled by the Monte Carlo Sampling (MCS) method to generate confidence intervals for the LES-computed velocity field. Also, Sobol indices were used to quantify the relative influence of each input parameter on the uncertainty of the LES results. This study revealed that uncertainty throughout most of the flow domain is primarily influenced by the flow discharge; however, near the banks and channel bed, bed roughness also contributed significantly to the uncertainty in the results.