In the research herein we focus on estimating the magnitude of scour at bridge sites using three different approaches: the SC Scour Envelope Curves, four simple analytical models in literature including HEC-18, and the use of advanced numerical model (Delft3D.) Field work is performed at four bridge sites and measurements of flow, bathymetry, and geometry data are obtained for calibration of the numerical modeling approach. Results from the various approaches are compared to provide the NCDOT with model performance data that are focused on assessing the magnitude of scour at bridge sites. Five scour analytical prediction models are assessed in terms of two statistical parameters, termed Mean Absolute Percentage Error (MAPE,) as a measure of accuracy of the prediction, and “level of conservatism” (as a measure of percent of cases where scour estimates from a given approach exceeded measured values.) Analyses performed using clear water data collected by Benedict and Caldwell (2014) indicated values of MAPE ranged from 238% to 336%, whereas conservatism ranged from 77% to 97.8%. Briaud (2014) model presented the least error (238%), and SC Envelope (2018) model presented the highest error (336%). The use of the SC Envelope model provided the most conservative scour depth (97.8%), whereas Briaud (2014) model presented the least conservatism (77%). In parallel when the live bed laboratory data collected in Benedict and Caldwell (2014) were used, values of MAPE ranged from 23.5% to 1218% with the SC envelope approach yielding the higher end of the MAPE, whereas conservatism ranged from 28.4% to 100%. The accuracy and conservatism of a specific model can however be adjusted by multiplying the scour depth by proposed modification factors. Focusing on the Roanoke River bridge site, the scour prediction from the deterministic approaches provided a scour depth in the range of 2.37 m to 3.24 m. These values are one order of magnitude higher than the scour estimated from the Delft 3D numerical approach. The scour estimates from Delft 3D were however obtained using the temporal hydrograph of 100-year flood event (where the water depth and velocities varied over a period of 30 days), while the scour estimates from the analytical model equations provide pier scour at equilibrium, assuming the water depth and velocities corresponding to the 100-year flood at the structure. A relationship between reliability index and scour factors is introduced herein. These “scour factors” are proposed for application to the deterministic scour predictions to account for uncertainty and inherent model bias.
