AUTOMATIC CALIBRATION OF A WAVE MODEL WITH AN EVOLUTIONARY BAYESIAN METHOD

Abstract

Bayesian Inference has been widely applied with success in science and engineering. One of its main uses is the inference of model parameters in order to reconcile model outputs with evidence provided by measures. In this article we propose this application for coastal engineering problems. Specifically, it is proposed to infer the parameters of a numerical wave model used to downscale wave reanalysis data to a coastal site. The proposed method is applied to a case study on the Uruguayan Atlantic coast, where a few month wave measure data series is available and needs to be extended in order to be used on an engineering project. The wave model used is SWAN and the data in deep waters and the wind data were obtained from the ERA-Interim reanalysis. At first, the method was tested with one and two parameters, since in these cases it is possible to compare the obtained results with a plot of the target function. Finally it was used to calibrate four parameters of the wave model and assess the uncertainty introduced by the selection of a set of parameters.