Headland-bay beach planform and tidal range: A neural network model

Abstract

An artificial neural network model of headland-bay beach planform is presented and developed. It differs from conventional models in two fundamental respects. First, the knowledge of the problem is not input to the model in the form of a given planform equation, but is acquired by the model itself from a series of examples of headland-bay beaches in the course of a training process—in other words, the model learns. Second, the new model takes into account tidal range, a significant factor influencing sediment transport. The artificial neural network is a multilayer perceptron trained with the Bayesian regularisation algorithm. For the selection of the optimum network architecture a thorough experimental study involving 240 neural networks is carried out. The model with the selected network architecture is trained and validated on a data set comprising 29 headland-bay beaches, and shown to outperform a conventional headland-bay beach planform model. The model is then applied to investigate the relation between tidal range and headland-bay beach planform. For this purpose the non-dimensional planforms corresponding to two values of the maximum tidal range ( T r = 0.5 m and 4.5 m) under different wave obliquities ( q 0 = 40°, 50° and 60°) are simulated. For the smallest wave obliquity (40°) the shoreline curves for both tidal ranges practically overlap. For greater wave obliquities the larger tidal range is found to produce a greater bay indentation, the difference increasing with wave obliquity. These results improve our understanding of headland-bay beach morphology and, in particular, of the manner in which it varies between meso-macrotidal and microtidal conditions due to the tidal modulation of the sediment transport vectors or its absence, respectively.