This thesis investigates the application of complex adaptive systems approaches (e.g. Artificial Neural Networks and Evolutionary Computation) to the study of coastal hydrodynamic and morphodynamic behaviour. Traditionally, nearshore morphological coastal system studies have developed an understanding of those physical processes occurring on both short temporal, and small spatial scales with a large degree of success. The associated approaches and concepts used to study the coastal system at these scales have primarily been linear in nature. However, when these approaches to studying the coastal system are extended to investigating larger temporal and spatial scales, which are commensurate with the aims of coastal management, results have had less success. The lack of success in developing an understanding of large scale coastal behaviour is to a large extent attributable to the complex behaviour associated with the coastal system. This complexity arises as a result of both the stochastic and chaotic nature of the coastal system. This allows small scale system understanding to be acquired but prevents the larger scale behaviour to be predicted effectively. This thesis presents four hydro-morphodynamic case studies to demonstrate the utility of complex adaptive system approaches for studying coastal systems. The first two demonstrate the application of Artificial Neural Networks, whilst the latter two illustrate the application of Evolutionary Computation. Case Study #1 considers the nature of the discrepancy between the observed location of wave breaking patterns over submerged sandbars and the actual sandbar locations. Artificial Neural Networks were able to quantitatively correct the observed locations to produce reliable estimates of the actual sand bar locations. Case Study #2 considers the development of an approach for the discrimination of shoreline location in video images for the production of intertidal maps of the nearshore region. In this case the system modelled by the Artificial Neural Network is the nature of the discrimination model carried out by the eye in delineating a shoreline feature between regions of sand and water. The Artificial Neural Network approach was shown to robustly recognise a range of shoreline features at a variety of beaches and hydrodynamic settings. Case Study #3 was the only purely hydrodynamic study considered in the thesis. It investigated the use of Evolutionary Computation to provide means of developing a parametric description of directional wave spectra in both reflective and nonreflective conditions. It is shown to provide a unifying approach which produces results which surpassed those achieved by traditional analysis approaches even though this may not strictly have been considered as a fiddly complex system. Case Study #4 is the most ambitious application and addresses the need for data reduction as a precursor when trying to study large scale morphodynamic data sets. It utilises Evolutionary Computation approaches to extract the significant morphodynamic variability evidenced in both directly and remotely sampled nearshore morphologies. Significant data reduction is achieved whilst reWning up to 90% of the original variability in the data sets. These case studies clearly demonstrate the ability of complex adaptive systems to be successfully applied to coastal system studies. This success has been shown to equal and sometimess surpass the results that may be obtained by traditional approaches. The strong performance of Complex Adaptive System approaches is closely linked to the level of complexity or non-linearity of the system being studied. Based on a qualitative evaluation, Evolutionary Computation was shown to demonstrate an advantage over Artificial Neural Networks in terms of the level of new insights which may be obtained. However, utility also needs to consider general ease of applicability and ease of implementation of the study approach. In this sense, Artificial Neural Networks demonstrate more utility for the study of coastal systems. The qualitative assessment approach used to evaluate the case studies in this thesis, may be used as a guide for choosing the appropriateness of either Artificial Neural Networks or Evolutionary Computation for future coastal system studies.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:268666 |
| Date | January 2003 |
| Creators | Kingston, Kenneth Samuel |
| Publisher | University of Plymouth |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | http://hdl.handle.net/10026.1/474 |
Page generated in 0.0015 seconds