BAYONET GPE: A generic and practical overtopping model that includes uncertainty

Mean wave overtopping discharge (q = l/s/m) is generally accepted to be a primary design criterion for assessing the performance of coastal structures, and is a boundary condition for many flood risk assessments. Modern methods for assessing wave overtopping discharges and its consequences have been discussed by Pullen et al (2007), and more recently updated in EurOtop II (2016). Amongst the various tools available for assessing wave overtopping, the use of Artificial Neural Networks (ANNs) has become increasingly popular and convenient to use since the original models of Van-Gent et al, 2007 and Kingston et al , 2008, and more recently by Formentin et al, 2017 and Zanuttigh et al, 2016. This paper introduces the next stage in the development of these models, and introduces the BAYONET Gaussian Process Emulation (BGPE) model. Using the same source data, this new method reduces uncertainties and gives clear guidance on the range and validity of the outputs.