Can the application of a multi-objective evolutionary algorithm improve conveyance estimation?

The following work explores the possibility of applying an evolutionary algorithm (NSGA-II) to automate the calibration procedure of a quasi 2-D numerical model. Seven rivers are examined, with discharges and flow depths ranging from 6 to 450 m3/s and 0.49 to 3.69 m, respectively. The calibration procedure is based on matching lateral distributions of depth-mean velocity, for which data exist. The results are compared with simulations undertaken by an expert user using traditional calibration techniques. The new approach compares favourably with the expert user and is able to provide an insight into the optimum calibration parameters, from which further scientific knowledge could be developed.