Sediment Modeling Panel (part 2): Sediment Model Calibration and Evaluation

Thomas and Cheng (2006)* famously claimed an uncalibrated sediment simulation is not a “model.” They assert that uncalibrated sediment simulations can only aspire to be “numerical experiments”; the uncertainty associated with sediment equations and data mean you do not have a “model” of a sediment system until you have reproduced system processes.

However, calibrating morphological models can be difficult and the calibration can take many forms. Sediment models have many uncertain variables/parameters modelers can “tune” to change their results. Sediment data are often noisy or inconsistent. Sediment modelers can choose from several possible calibration targets to evaluate their model. Variables calibrated over one time series (e.g. a low-to-moderate flow period) might not hold up over another time-window (e.g. a flood). Equifinality poses possibilities of offsetting errors that generate the right answer for the wrong reasons, resulting in a calibrated model that is not general or predictive. Even the language surrounding sediment model evaluation (e.g. calibration, validation, verification, circumstantiation) can get confusing.

This panel discussion will interrogate four, federal, morphological model developers (representing SRH 1D, HEC-RAS 1D&2D and ADH) about their model evaluation practices and reconditions.

The panel will discuss the following topics:

• The importance of morphological model calibration. • The stages of a morphological model calibration o (e.g. steady flow-stable bed, unsteady flow-stable bed, steady flow-mobile bed, unsteady flow-mobile bed). • Aligning calibration with modeling objectives. • Model evaluation targets (i.e. calibration metrics). • Multi-time series and multi-variable model evaluation. • Selecting calibration parameters/variables (i.e. what do we change). • Level of effort to expect/scope for calibration

*ASCE Manual of Practice 110 – Sedimentation Engineering