Instantaneous energetics sediment transport model calibration

In this note we point out a bias error that affects calibrations of 'Bagnold-type' energetics sediment transport models. Calibrations based on instantaneous measurements of fluid velocity and suspended sediment concentration incur an inherent increase in correlation between measured and predicted sediment transport rates because the measured fluid velocity resides on both sides of the calibration equation. Random, fully uncorrelated velocity and suspended load time series tests comparing the energetics model with a similar model which divides both sides of the equation by the velocity, showed that having velocity on both sides increased the R2 correlation coefficient from its expected near zero value to 0.45. This “false correlation” can be as high as 0.55 when there is a high mean concentration relative to the concentration variance and there are small mean velocities. In contrast, when there is relatively high variability in concentration in the presence of large mean velocities (e.g. suspension events of coarse grains under waves in the surf zone with an alongshore current), the “false correlation” reduced to 0.35. Comparisons with data from two swash experiments and a surf zone study showed a similar increase in “false correlation”. Associated with the “false correlation” was a 4-fold overestimate of the calibration coefficient used to tune the sediment transport model under simulated noisy field measurement conditions.