Internal testing of a numerical model of hillslope-channel coupling using laboratory flume experiments

The numerical model COUP2D simulates the hydrological coupling between hillslopes and the river channel during a rainfall event. In order to test the numerical model, a 1:100 scaled laboratory flume which was modified to incorporate lateral hillslope elements, was used to run a series of experiments in which hillslope angle, channel angle, hillslope discharge and channel discharge were the varying parameters. Overall, there were 18 different experimental configurations with three replicates carried out for each condition, leading to a total of 54 experiments. These conditions were then used to parameterize and run COUP2D. Internal model outputs of flow depth and flow velocity at four cross-sections in the channel were compared to the measurements made in the physical model for the same parameter conditions. Statistical comparisons of the measured and modelled data were carried out for each experiment and across all experiments, using two goodness-of-fit measures - root mean square error and Nash-Sutcliffe coefficient of efficiency - in order to assess the performance of the model over an entire simulation as well as over all the simulations. The main effects on the goodness-of-fit measures for flow depth of each experimental variable, as well as the interactions between variables, were evaluated using statistical modelling. The results show that the model captures flow-depth variations in response to changing channel and hillslope parameters. Statistical modelling suggests that the main effects on model error are cross-section position, channel angle and channel discharge. Significant interactions also occur between all the channel variables and between the channel variables and hillslope discharge. The results of the testing procedure have significant implications for the consideration of different model components and for the interaction between data- and model evaluation.